diff options
Diffstat (limited to 'libbpkg/manifest.cxx')
-rw-r--r-- | libbpkg/manifest.cxx | 4348 |
1 files changed, 3821 insertions, 527 deletions
diff --git a/libbpkg/manifest.cxx b/libbpkg/manifest.cxx index 8631d18..bd69b85 100644 --- a/libbpkg/manifest.cxx +++ b/libbpkg/manifest.cxx @@ -4,24 +4,32 @@ #include <libbpkg/manifest.hxx> #include <string> +#include <limits> #include <ostream> #include <sstream> #include <cassert> -#include <cstring> // strncmp(), strcmp() -#include <utility> // move() -#include <cstdint> // uint*_t, UINT16_MAX -#include <algorithm> // find(), find_if_not(), find_first_of(), replace() -#include <stdexcept> // invalid_argument - -#include <libbutl/url.mxx> -#include <libbutl/path.mxx> -#include <libbutl/base64.mxx> -#include <libbutl/utility.mxx> // icasecmp(), lcase(), alnum(), +#include <cstdlib> // strtoull() +#include <cstring> // strncmp(), strcmp(), strchr(), strcspn() +#include <utility> // move() +#include <cstdint> // uint*_t +#include <algorithm> // find(), find_if(), find_first_of(), replace() +#include <stdexcept> // invalid_argument +#include <type_traits> // remove_reference + +#include <libbutl/url.hxx> +#include <libbutl/path.hxx> +#include <libbutl/utf8.hxx> +#include <libbutl/base64.hxx> +#include <libbutl/utility.hxx> // icasecmp(), lcase(), alnum(), // digit(), xdigit(), next_word() -#include <libbutl/small-vector.mxx> -#include <libbutl/manifest-parser.mxx> -#include <libbutl/manifest-serializer.mxx> -#include <libbutl/standard-version.mxx> +#include <libbutl/filesystem.hxx> // dir_exist() +#include <libbutl/small-vector.hxx> +#include <libbutl/char-scanner.hxx> +#include <libbutl/manifest-parser.hxx> +#include <libbutl/manifest-serializer.hxx> +#include <libbutl/standard-version.hxx> + +#include <libbpkg/buildfile-scanner.hxx> using namespace std; using namespace butl; @@ -171,12 +179,12 @@ namespace bpkg canonical_upstream ( data_type (upstream.c_str (), data_type::parse::upstream, - false /* fold_zero_revision */). + none). canonical_upstream), canonical_release ( data_type (release ? release->c_str () : nullptr, data_type::parse::release, - false /* fold_zero_revision */). + none). canonical_release) { // Check members constrains. @@ -253,9 +261,12 @@ namespace bpkg } version::data_type:: - data_type (const char* v, parse pr, bool fold_zero_rev) + data_type (const char* v, parse pr, version::flags fl) { - if (fold_zero_rev) + if ((fl & version::fold_zero_revision) != 0) + assert (pr == parse::full); + + if ((fl & version::allow_iteration) != 0) assert (pr == parse::full); // Otherwise compiler gets confused with string() member. @@ -270,32 +281,75 @@ namespace bpkg return; } - assert (v != nullptr); - - optional<uint16_t> ep; - auto bad_arg = [](const string& d) {throw invalid_argument (d);}; - auto uint16 = [&bad_arg](const string& s, const char* what) -> uint16_t + auto parse_uint = [&bad_arg](const string& s, auto& r, const char* what) { - try - { - uint64_t v (stoull (s)); + using type = typename remove_reference<decltype (r)>::type; - if (v <= UINT16_MAX) // From <cstdint>. - return static_cast<uint16_t> (v); - } - catch (const std::exception&) + if (!s.empty () && s[0] != '-' && s[0] != '+') // strtoull() allows these. { - // Fall through. + const char* b (s.c_str ()); + char* e (nullptr); + errno = 0; // We must clear it according to POSIX. + uint64_t v (strtoull (b, &e, 10)); // Can't throw. + + if (errno != ERANGE && + e == b + s.size () && + v <= numeric_limits<type>::max ()) + { + r = static_cast<type> (v); + return; + } } - bad_arg (string (what) + " should be 2-byte unsigned integer"); + bad_arg (string (what) + " should be " + + std::to_string (sizeof (type)) + "-byte unsigned integer"); + }; - assert (false); // Can't be here. - return 0; + auto parse_uint16 = [&parse_uint](const string& s, const char* what) + { + uint16_t r; + parse_uint (s, r, what); + return r; }; + auto parse_uint32 = [&parse_uint](const string& s, const char* what) + { + uint32_t r; + parse_uint (s, r, what); + return r; + }; + + assert (v != nullptr); + + // Parse the iteration, if allowed. + // + // Note that allowing iteration is not very common, so let's handle it in + // an ad hoc way not to complicate the subsequent parsing. + // + string storage; + if (pr == parse::full) + { + iteration = 0; + + // Note that if not allowed but the iteration is present, then the below + // version parsing code will fail with appropriate diagnostics. + // + if ((fl & version::allow_iteration) != 0) + { + if (const char* p = strchr (v, '#')) + { + iteration = parse_uint32 (p + 1, "iteration"); + + storage.assign (v, p - v); + v = storage.c_str (); + } + } + } + + optional<uint16_t> ep; + enum class mode {epoch, upstream, release, revision}; mode m (pr == parse::full ? (v[0] == '+' @@ -350,7 +404,7 @@ namespace bpkg if (lnn >= cb) // Contains non-digits. bad_arg ("epoch should be 2-byte unsigned integer"); - ep = uint16 (string (cb, p), "epoch"); + ep = parse_uint16 (string (cb, p), "epoch"); } else canon_part->add (cb, p, lnn < cb); @@ -423,9 +477,9 @@ namespace bpkg if (lnn >= cb) // Contains non-digits. bad_arg ("revision should be 2-byte unsigned integer"); - std::uint16_t rev (uint16 (cb, "revision")); + uint16_t rev (parse_uint16 (cb, "revision")); - if (rev != 0 || !fold_zero_rev) + if (rev != 0 || (fl & fold_zero_revision) == 0) revision = rev; } else if (cb != p) @@ -512,7 +566,7 @@ namespace bpkg } version& version:: - operator= (version&& v) + operator= (version&& v) noexcept { if (this != &v) { @@ -570,7 +624,8 @@ namespace bpkg } text_file:: - text_file (text_file&& f): file (f.file), comment (move (f.comment)) + text_file (text_file&& f) noexcept + : file (f.file), comment (move (f.comment)) { if (file) new (&path) path_type (move (f.path)); @@ -588,12 +643,12 @@ namespace bpkg } text_file& text_file:: - operator= (text_file&& f) + operator= (text_file&& f) noexcept { if (this != &f) { this->~text_file (); - new (this) text_file (move (f)); // Assume noexcept move-construction. + new (this) text_file (move (f)); // Rely on noexcept move-construction. } return *this; } @@ -606,6 +661,132 @@ namespace bpkg return *this; } + // text_type + // + string + to_string (text_type t) + { + switch (t) + { + case text_type::plain: return "text/plain"; + case text_type::github_mark: return "text/markdown;variant=GFM"; + case text_type::common_mark: return "text/markdown;variant=CommonMark"; + } + + assert (false); // Can't be here. + return string (); + } + + optional<text_type> + to_text_type (const string& t) + { + auto bad_type = [] (const string& d) {throw invalid_argument (d);}; + + // Parse the media type representation (see RFC2045 for details) into the + // type/subtype value and the parameter list. Note: we don't support + // parameter quoting and comments for simplicity. + // + size_t p (t.find (';')); + const string& tp (p != string::npos ? trim (string (t, 0, p)) : t); + + small_vector<pair<string, string>, 1> ps; + + while (p != string::npos) + { + // Extract parameter name. + // + size_t b (p + 1); + p = t.find ('=', b); + + if (p == string::npos) + bad_type ("missing '='"); + + string n (trim (string (t, b, p - b))); + + // Extract parameter value. + // + b = p + 1; + p = t.find (';', b); + + string v (trim (string (t, + b, + p != string::npos ? p - b : string::npos))); + + ps.emplace_back (move (n), move (v)); + } + + // Calculate the resulting text type, failing on unrecognized media type, + // unexpected parameter name or value. + // + // Note that type, subtype, and parameter names are matched + // case-insensitively. + // + optional<text_type> r; + + // Currently only the plain and markdown text types are allowed. Later we + // can potentially introduce some other text types. + // + if (icasecmp (tp, "text/plain") == 0) + { + // Currently, we don't expect parameters for plain text. Later we can + // potentially introduce some plain text variants. + // + if (ps.empty ()) + r = text_type::plain; + } + else if (icasecmp (tp, "text/markdown") == 0) + { + // Currently, a single optional variant parameter with the two possible + // values is allowed for markdown. Later we can potentially introduce + // some other markdown variants. + // + if (ps.empty () || + (ps.size () == 1 && icasecmp (ps[0].first, "variant") == 0)) + { + // Note that markdown variants are matched case-insensitively (see + // RFC7763 for details). + // + string v; + if (ps.empty () || icasecmp (v = move (ps[0].second), "GFM") == 0) + r = text_type::github_mark; + else if (icasecmp (v, "CommonMark") == 0) + r = text_type::common_mark; + } + } + else if (icasecmp (tp, "text/", 5) != 0) + bad_type ("text type expected"); + + return r; + } + + // typed_text_file + // + optional<text_type> typed_text_file:: + effective_type (bool iu) const + { + optional<text_type> r; + + if (type) + { + r = to_text_type (*type); + } + else if (file) + { + string ext (path.extension ()); + if (ext.empty () || icasecmp (ext, "txt") == 0) + r = text_type::plain; + else if (icasecmp (ext, "md") == 0 || icasecmp (ext, "markdown") == 0) + r = text_type::github_mark; + } + else + r = text_type::plain; + + if (!r && !iu) + throw invalid_argument ("unknown text type"); + + return r; + } + // manifest_url // manifest_url:: @@ -657,7 +838,7 @@ namespace bpkg if (mnv != "$") try { - min_version = version (mnv, false /* fold_zero_revision */); + min_version = version (mnv, version::none); } catch (const invalid_argument& e) { @@ -684,7 +865,7 @@ namespace bpkg if (mxv != "$") try { - max_version = version (mxv, false /* fold_zero_revision */); + max_version = version (mxv, version::none); } catch (const invalid_argument& e) { @@ -788,7 +969,7 @@ namespace bpkg // version. // if (vs != "$") - v = version (vs, false /* fold_zero_revision */); + v = version (vs, version::none); switch (operation) { @@ -1012,15 +1193,146 @@ namespace bpkg return r; } - std::string dependency:: + // dependency + // + dependency:: + dependency (std::string d) + { + using std::string; + using iterator = string::const_iterator; + + iterator b (d.begin ()); + iterator i (b); + iterator ne (b); // End of name. + iterator e (d.end ()); + + // Find end of name (ne). + // + // Grep for '=<>([~^' in the bpkg source code and update, if changed. + // + const string cb ("=<>([~^"); + for (char c; i != e && cb.find (c = *i) == string::npos; ++i) + { + if (!space (c)) + ne = i + 1; + } + + try + { + name = package_name (i == e ? move (d) : string (b, ne)); + } + catch (const invalid_argument& e) + { + throw invalid_argument (string ("invalid package name: ") + e.what ()); + } + + if (i != e) + try + { + constraint = version_constraint (string (i, e)); + } + catch (const invalid_argument& e) + { + throw invalid_argument (string ("invalid package constraint: ") + + e.what ()); + } + } + + // dependency_alternative + // + string dependency_alternative:: string () const { - std::string r (name.string ()); + std::string r (size () > 1 ? "{" : ""); - if (constraint) + bool first (true); + for (const dependency& d: *this) { - r += ' '; - r += constraint->string (); + if (!first) + r += ' '; + else + first = false; + + r += d.string (); + } + + if (size () > 1) + r += '}'; + + if (single_line ()) + { + if (enable) + { + r += " ? ("; + r += *enable; + r += ')'; + } + + if (reflect) + { + r += ' '; + r += *reflect; + } + } + else + { + // Add an extra newline between the clauses. + // + bool first (true); + + r += "\n{"; + + if (enable) + { + first = false; + + r += "\n enable ("; + r += *enable; + r += ')'; + } + + if (prefer) + { + if (!first) + r += '\n'; + else + first = false; + + r += "\n prefer\n {\n"; + r += *prefer; + r += " }"; + + assert (accept); + + r += "\n\n accept ("; + r += *accept; + r += ')'; + } + else if (require) + { + if (!first) + r += '\n'; + else + first = false; + + r += "\n require\n {\n"; + r += *require; + r += " }"; + } + + if (reflect) + { + if (!first) + r += '\n'; + else + first = false; + + r += "\n reflect\n {\n"; + r += *reflect; + r += " }"; + } + + r += "\n}"; } return r; @@ -1028,26 +1340,1370 @@ namespace bpkg // dependency_alternatives // - ostream& - operator<< (ostream& o, const dependency_alternatives& as) + class dependency_alternatives_lexer: public char_scanner<utf8_validator> + { + public: + enum class token_type + { + eos, + newline, + word, + buildfile, + + question, // ? + + lcbrace, // { + rcbrace, // } + + lparen, // ( + rparen, // ) + + lsbrace, // [ + rsbrace, // ] + + equal, // == + less, // < + greater, // > + less_equal, // <= + greater_equal, // >= + + tilde, // ~ + caret, // ^ + + bit_or // | + }; + + struct token + { + token_type type; + std::string value; + + uint64_t line; + uint64_t column; + + std::string + string (bool diag = true) const; + }; + + // If true, then comments are allowed and are treated as whitespace + // characters. + // + bool comments = false; + + public: + // Note that name is stored by shallow reference. + // + dependency_alternatives_lexer (istream& is, + const string& name, + uint64_t line, + uint64_t column) + : char_scanner (is, + utf8_validator (codepoint_types::graphic, U"\n\r\t"), + true /* crlf */, + line, + column), + name_ (name), + buildfile_scan_ (*this, name_) {} + + // The following functions throw manifest_parsing on invalid UTF-8 + // sequence. + // + + // Peek the next non-whitespace character. + // + xchar + peek_char (); + + // Extract next token (other than of the buildfile type) from the stream. + // + // Note that it is ok to call next() again after getting eos. + // + token + next (); + + // The next_*() functions extract the buildfile token from the stream. + // Throw manifest_parsing on error (invalid buildfile fragment, etc). + // + // Note that they are just thin wrappers around the scan_*() functions + // (see buildfile-scanner.hxx for details). + // + token + next_eval (); + + token + next_line (char stop); + + token + next_block (); + + private: + using base = char_scanner<utf8_validator>; + + xchar + get () + { + xchar c (base::get (ebuf_)); + + if (invalid (c)) + throw parsing (name_, c.line, c.column, ebuf_); + + return c; + } + + void + get (const xchar& peeked) + { + base::get (peeked); + } + + xchar + peek () + { + xchar c (base::peek (ebuf_)); + + if (invalid (c)) + throw parsing (name_, c.line, c.column, ebuf_); + + return c; + } + + void + skip_spaces (); + + private: + const string& name_; + + // Buffer for a get()/peek() potential error. + // + string ebuf_; + + buildfile_scanner<utf8_validator, 1> buildfile_scan_; + }; + + dependency_alternatives_lexer::token dependency_alternatives_lexer:: + next () + { + using type = token_type; + + skip_spaces (); + + uint64_t ln (line); + uint64_t cl (column); + + xchar c (get ()); + + auto make_token = [ln, cl] (type t, string v = string ()) + { + return token {t, move (v), ln, cl}; + }; + + if (eos (c)) + return make_token (type::eos); + + // NOTE: don't forget to also update the below separators list if changing + // anything here. + // + switch (c) + { + case '\n': return make_token (type::newline); + case '?': return make_token (type::question); + case '(': return make_token (type::lparen); + case ')': return make_token (type::rparen); + case '{': return make_token (type::lcbrace); + case '}': return make_token (type::rcbrace); + case '[': return make_token (type::lsbrace); + case ']': return make_token (type::rsbrace); + + case '=': + { + if (peek () == '=') + { + get (); + return make_token (type::equal); + } + break; + } + + case '<': + { + if ((c = peek ()) == '=') + { + get (c); + return make_token (type::less_equal); + } + else + return make_token (type::less); + } + + case '>': + { + if ((c = peek ()) == '=') + { + get (c); + return make_token (type::greater_equal); + } + else + return make_token (type::greater); + } + + case '~': return make_token (type::tilde); + case '^': return make_token (type::caret); + + case '|': return make_token (type::bit_or); + } + + // Otherwise it is a word. + // + // Starts with a non-whitespace character which has not been recognized as + // a part of some other token. + // + string r (1, c); + + // Add subsequent characters until eos or separator is encountered. + // + const char* s (" \n\t?(){}[]=<>~^|"); + for (c = peek (); !eos (c) && strchr (s, c) == nullptr; c = peek ()) + { + r += c; + get (c); + } + + return make_token (type::word, move (r)); + } + + dependency_alternatives_lexer::token dependency_alternatives_lexer:: + next_eval () + { + skip_spaces (); + + uint64_t ln (line); + uint64_t cl (column); + + try + { + // Strip the trailing whitespaces. + // + return token {token_type::buildfile, + trim (buildfile_scan_.scan_eval ()), + ln, + cl}; + } + catch (const buildfile_scanning& e) + { + throw parsing (e.name, e.line, e.column, e.description); + } + } + + dependency_alternatives_lexer::token dependency_alternatives_lexer:: + next_line (char stop) + { + skip_spaces (); + + uint64_t ln (line); + uint64_t cl (column); + + try + { + // Strip the trailing whitespaces. + // + return token {token_type::buildfile, + trim (buildfile_scan_.scan_line (stop)), + ln, + cl}; + } + catch (const buildfile_scanning& e) + { + throw parsing (e.name, e.line, e.column, e.description); + } + } + + dependency_alternatives_lexer::token dependency_alternatives_lexer:: + next_block () + { + uint64_t ln (line); + uint64_t cl (column); + + try + { + // Don't trim the token value not to strip the potential block indenting + // on the first line. + // + return token {token_type::buildfile, + buildfile_scan_.scan_block (), + ln, + cl}; + } + catch (const buildfile_scanning& e) + { + throw parsing (e.name, e.line, e.column, e.description); + } + } + + dependency_alternatives_lexer::xchar dependency_alternatives_lexer:: + peek_char () + { + skip_spaces (); + return peek (); + } + + void dependency_alternatives_lexer:: + skip_spaces () + { + xchar c (peek ()); + bool start (c.column == 1); + + for (; !eos (c); c = peek ()) + { + switch (c) + { + case ' ': + case '\t': break; + + case '#': + { + if (!comments) + return; + + get (c); + + // See if this is a multi-line comment in the form: + // + /* + #\ + ... + #\ + */ + auto ml = [&c, this] () -> bool + { + if ((c = peek ()) == '\\') + { + get (c); + if ((c = peek ()) == '\n' || eos (c)) + return true; + } + + return false; + }; + + if (ml ()) + { + // Scan until we see the closing one. + // + for (;;) + { + if (c == '#' && ml ()) + break; + + if (eos (c = peek ())) + throw parsing (name_, + c.line, c.column, + "unterminated multi-line comment"); + + get (c); + } + } + else + { + // Read until newline or eos. + // + for (; !eos (c) && c != '\n'; c = peek ()) + get (c); + } + + continue; + } + + case '\n': + { + // Skip empty lines. + // + if (start) + break; + } + // Fall through. + default: return; + } + + get (c); + } + } + + std::string dependency_alternatives_lexer::token:: + string (bool diag) const + { + std::string q (diag ? "'" : ""); + + switch (type) + { + case token_type::eos: return diag ? "<end of stream>" : ""; + case token_type::newline: return diag ? "<newline>" : "\n"; + case token_type::word: return q + value + q; + case token_type::buildfile: return (diag + ? "<buildfile fragment>" + : value); + case token_type::question: return q + '?' + q; + case token_type::lparen: return q + '(' + q; + case token_type::rparen: return q + ')' + q; + case token_type::lcbrace: return q + '{' + q; + case token_type::rcbrace: return q + '}' + q; + case token_type::lsbrace: return q + '[' + q; + case token_type::rsbrace: return q + ']' + q; + case token_type::equal: return q + "==" + q; + case token_type::less: return q + '<' + q; + case token_type::greater: return q + '>' + q; + case token_type::less_equal: return q + "<=" + q; + case token_type::greater_equal: return q + ">=" + q; + case token_type::tilde: return q + '~' + q; + case token_type::caret: return q + '^' + q; + case token_type::bit_or: return q + '|' + q; + } + + assert (false); // Can't be here. + return ""; + } + + class dependency_alternatives_parser + { + public: + + // If the requirements flavor is specified, then only enable and reflect + // clauses are allowed in the multi-line representation. + // + explicit + dependency_alternatives_parser (bool requirements = false) + : requirements_ (requirements) {} + + // Throw manifest_parsing if representation is invalid. + // + void + parse (const package_name& dependent, + istream&, + const string& name, + uint64_t line, + uint64_t column, + dependency_alternatives&); + + private: + using lexer = dependency_alternatives_lexer; + using token = lexer::token; + using token_type = lexer::token_type; + + token_type + next (token&, token_type&); + + token_type + next_eval (token&, token_type&); + + token_type + next_line (token&, token_type&); + + token_type + next_block (token&, token_type&); + + // Receive the token/type from which it should start consuming and in + // return the token/type contains the first token that has not been + // consumed (normally eos, newline, or '|'). + // + dependency_alternative + parse_alternative (token&, token_type&, bool first); + + // Helpers. + // + // Throw manifest_parsing with the `<what> expected instead of <token>` + // description. + // + [[noreturn]] void + unexpected_token (const token&, string&& what); + + bool requirements_; + + const package_name* dependent_; + const string* name_; + lexer* lexer_; + dependency_alternatives* result_; + }; + + [[noreturn]] void dependency_alternatives_parser:: + unexpected_token (const token& t, string&& w) + { + w += " expected"; + + // Don't add the `instead of...` part, if the unexpected token is eos or + // an empty word/buildfile. + // + if (t.type != token_type::eos && + ((t.type != token_type::word && t.type != token_type::buildfile) || + !t.value.empty ())) + { + w += " instead of "; + w += t.string (); + } + + throw parsing (*name_, t.line, t.column, w); + } + + void dependency_alternatives_parser:: + parse (const package_name& dependent, + istream& is, + const string& name, + uint64_t line, + uint64_t column, + dependency_alternatives& result) + { + lexer lexer (is, name, line, column); + + dependent_ = &dependent; + name_ = &name; + lexer_ = &lexer; + result_ = &result; + + string what (requirements_ ? "requirement" : "dependency"); + + token t; + token_type tt; + next (t, tt); + + // Make sure the representation is not empty, unless we are in the + // requirements mode. In the latter case fallback to creating a simple + // unconditional requirement. Note that it's the caller's responsibility + // to verify that a non-empty comment is specified in this case. + // + if (tt == token_type::eos) + { + if (!requirements_) + unexpected_token (t, what + " alternatives"); + + dependency_alternative da; + da.push_back (dependency ()); + + result_->push_back (move (da)); + return; + } + + for (bool first (true); tt != token_type::eos; ) + { + dependency_alternative da (parse_alternative (t, tt, first)); + + // Skip newline after the dependency alternative, if present. + // + if (tt == token_type::newline) + next (t, tt); + + // Make sure that the simple requirement has the only alternative in the + // representation. + // + if (requirements_ && + da.size () == 1 && + (da[0].name.empty () || (da.enable && da.enable->empty ()))) + { + assert (first); + + if (tt != token_type::eos) + throw parsing (*name_, + t.line, + t.column, + "end of simple requirement expected"); + } + else + { + if (tt != token_type::eos && tt != token_type::bit_or) + unexpected_token (t, "end of " + what + " alternatives or '|'"); + } + + if (tt == token_type::bit_or) + { + next (t, tt); + + // Skip newline after '|', if present. + // + if (tt == token_type::newline) + next (t, tt); + + // Make sure '|' is not followed by eos. + // + if (tt == token_type::eos) + unexpected_token (t, move (what)); + } + + result_->push_back (move (da)); + + first = false; + } + } + + dependency_alternative dependency_alternatives_parser:: + parse_alternative (token& t, token_type& tt, bool first) + { + using type = token_type; + + dependency_alternative r; + + string what (requirements_ ? "requirement" : "dependency"); + string config ("config." + dependent_->variable () + '.'); + + auto bad_token = [&t, this] (string&& what) + { + unexpected_token (t, move (what)); + }; + + // Check that the current token type matches the expected one. Throw + // manifest_parsing if that's not the case. Use the expected token type + // name for the error description or the custom name, if specified. For + // the word and buildfile token types the custom name must be specified. + // + // Only move from the custom name argument if throwing exception. + // + auto expect_token = [&tt, &bad_token] (type et, + string&& what = string ()) + { + assert ((et != type::word && et != type::buildfile) || !what.empty ()); + + if (tt != et) + { + if (what.empty ()) + { + token e {et, "", 0, 0}; + bad_token (e.string ()); + } + else + bad_token (move (what)); + } + }; + + // Parse dependencies. + // + // If the current token starts the version constraint, then read its + // tokens, rejoin them, and return the constraint string representation. + // Otherwise return nullopt. + // + // Note that normally the caller reads the dependency package name, reads + // the version constraint and, if present, appends it to the dependency, + // and then creates the dependency object with a single constructor call. + // + // Note: doesn't read token that follows the constraint. + // + auto try_scan_version_constraint = + [&t, &tt, &bad_token, &expect_token, this] () -> optional<string> + { + switch (t.type) + { + case type::lparen: + case type::lsbrace: + { + string r (t.string (false /* diag */)); + + next (t, tt); + + expect_token (type::word, "version"); + + r += t.string (false /* diag */); + r += ' '; + + next (t, tt); + + expect_token (type::word, "version"); + + r += t.string (false /* diag */); + + next (t, tt); + + if (tt != type::rparen && tt != type::rsbrace) + bad_token ("')' or ']'"); + + r += t.string (false /* diag */); + + return optional<string> (move (r)); + } + + case type::equal: + case type::less: + case type::greater: + case type::less_equal: + case type::greater_equal: + case type::tilde: + case type::caret: + { + string r (t.string (false /* diag */)); + + next (t, tt); + + expect_token (type::word, "version"); + + r += t.string (false /* diag */); + + return optional<string> (move (r)); + } + + default: return nullopt; + } + }; + + // Parse the evaluation context including the left and right parenthesis + // and return the enclosed buildfile fragment. + // + // Note: no token is read after terminating ')'. + // + auto parse_eval = [&t, &tt, &expect_token, &bad_token, this] () + { + next (t, tt); + expect_token (type::lparen); + + next_eval (t, tt); + + if (t.value.empty ()) + bad_token ("condition"); + + string r (move (t.value)); + + next (t, tt); + expect_token (type::rparen); + + return r; + }; + + const char* vccs ("([<>=!~^"); + + bool group (tt == type::lcbrace); // Dependency group. + + if (group) + { + next (t, tt); + + if (tt == type::rcbrace) + bad_token (move (what)); + + while (tt != type::rcbrace) + { + expect_token (type::word, what + " or '}'"); + + string d (move (t.value)); + uint64_t dl (t.line); + uint64_t dc (t.column); + + next (t, tt); + + optional<string> vc (try_scan_version_constraint ()); + + if (vc) + { + d += *vc; + + next (t, tt); + } + + try + { + r.emplace_back (d); + } + catch (const invalid_argument& e) + { + throw parsing (*name_, dl, dc, e.what ()); + } + } + + // See if a common version constraint follows the dependency group and + // parse it if that's the case. + // + // Note that we need to be accurate not to consume what may end up to be + // a part of the reflect config. + // + lexer::xchar c (lexer_->peek_char ()); + + if (!lexer::eos (c) && strchr (vccs, c) != nullptr) + { + next (t, tt); + + uint64_t vcl (t.line); + uint64_t vcc (t.column); + + optional<string> vc (try_scan_version_constraint ()); + + if (!vc) + bad_token ("version constraint"); + + try + { + version_constraint c (*vc); + + for (dependency& d: r) + { + if (!d.constraint) + d.constraint = c; + } + } + catch (const invalid_argument& e) + { + throw parsing (*name_, + vcl, + vcc, + string ("invalid version constraint: ") + e.what ()); + } + } + } + else // Single dependency. + { + // If we see the question mark instead of a word in the requirements + // mode, then this is a simple requirement. In this case parse the + // evaluation context, if present, and bail out. + // + if (requirements_ && first && tt == type::question) + { + r.emplace_back (dependency ()); + + bool eval (lexer_->peek_char () == '('); + r.enable = eval ? parse_eval () : string (); + + next (t, tt); + + // @@ TMP Treat requirements similar to `? cli` as `cli ?` until + // toolchain 0.15.0 and libodb-mssql 2.5.0-b.22 are both released. + // + // NOTE: don't forget to drop the temporary test in + // tests/manifest/testscript when dropping this workaround. + // + if (!eval && tt == type::word) + try + { + r.back ().name = package_name (move (t.value)); + next (t, tt); + } + catch (const invalid_argument&) {} + + return r; + } + + expect_token (type::word, move (what)); + + string d (move (t.value)); + uint64_t dl (t.line); + uint64_t dc (t.column); + + // See if a version constraint follows the dependency package name and + // parse it if that's the case. + // + lexer::xchar c (lexer_->peek_char ()); + + if (!lexer::eos (c) && strchr (vccs, c) != nullptr) + { + next (t, tt); + + optional<string> vc (try_scan_version_constraint ()); + + if (!vc) + bad_token ("version constraint"); + + d += *vc; + } + + try + { + r.emplace_back (d); + } + catch (const invalid_argument& e) + { + throw parsing (*name_, dl, dc, e.what ()); + } + } + + // See if there is an enable condition and parse it if that's the case. + // + { + lexer::xchar c (lexer_->peek_char ()); + + if (c == '?') + { + next (t, tt); + expect_token (type::question); + + // If we don't see the opening parenthesis in the requirements mode, + // then this is a simple requirement. In this case set the enable + // condition to an empty string and bail out. + // + c = lexer_->peek_char (); + + if (requirements_ && first && !group && c != '(') + { + r.enable = ""; + + next (t, tt); + return r; + } + + r.enable = parse_eval (); + } + } + + // See if there is a reflect config and parse it if that's the case. + // + { + lexer::xchar c (lexer_->peek_char ()); + + if (!lexer::eos (c) && strchr ("|\n", c) == nullptr) + { + next_line (t, tt); + + string& l (t.value); + if (l.compare (0, config.size (), config) != 0) + bad_token (config + "* variable assignment"); + + r.reflect = move (l); + } + } + + // If the dependencies are terminated with the newline, then check if the + // next token is '{'. If that's the case, then this is a multi-line + // representation. + // + next (t, tt); + + if (tt == type::newline) + { + next (t, tt); + + if (tt == type::lcbrace) + { + if (r.enable) + throw parsing ( + *name_, + t.line, + t.column, + "multi-line " + what + " form with inline enable clause"); + + if (r.reflect) + throw parsing ( + *name_, + t.line, + t.column, + "multi-line " + what + " form with inline reflect clause"); + + // Allow comments. + // + lexer_->comments = true; + + next (t, tt); + expect_token (type::newline); + + // Parse the clauses. + // + for (next (t, tt); tt == type::word; next (t, tt)) + { + auto fail_dup = [&t, this] () + { + throw parsing (*name_, t.line, t.column, "duplicate clause"); + }; + + auto fail_precede = [&t, this] (const char* what) + { + throw parsing ( + *name_, + t.line, + t.column, + t.value + " clause should precede " + what + " clause"); + }; + + auto fail_conflict = [&t, this] (const char* what) + { + throw parsing ( + *name_, + t.line, + t.column, + t.value + " and " + what + " clauses are mutually exclusive"); + }; + + auto fail_requirements = [&t, this] () + { + throw parsing ( + *name_, + t.line, + t.column, + t.value + " clause is not permitted for requirements"); + }; + + // Parse the buildfile fragment block including the left and right + // curly braces (expected to be on the separate lines) and return + // the enclosed fragment. + // + // Note that an empty buildfile fragment is allowed. + // + auto parse_block = [&t, &tt, &expect_token, this] () + { + next (t, tt); + expect_token (type::newline); + + next (t, tt); + expect_token (type::lcbrace); + + next (t, tt); + expect_token (type::newline); + + next_block (t, tt); + + return move (t.value); + }; + + const string& v (t.value); + + if (v == "enable") + { + if (r.enable) + fail_dup (); + + if (r.prefer) + fail_precede ("prefer"); + + if (r.require) + fail_precede ("require"); + + if (r.reflect) + fail_precede ("reflect"); + + r.enable = parse_eval (); + + next (t, tt); + expect_token (type::newline); + } + else if (v == "prefer") + { + if (requirements_) + fail_requirements (); + + if (r.prefer) + fail_dup (); + + if (r.require) + fail_conflict ("require"); + + if (r.reflect) + fail_precede ("reflect"); + + r.prefer = parse_block (); + + // The accept clause must follow, so parse it. + // + next (t, tt); + + if (tt != type::word || t.value != "accept") + bad_token ("accept clause"); + + r.accept = parse_eval (); + + next (t, tt); + expect_token (type::newline); + } + else if (v == "require") + { + if (requirements_) + fail_requirements (); + + if (r.require) + fail_dup (); + + if (r.prefer) + fail_conflict ("prefer"); + + if (r.reflect) + fail_precede ("reflect"); + + r.require = parse_block (); + } + else if (v == "reflect") + { + if (r.reflect) + fail_dup (); + + r.reflect = parse_block (); + } + else if (v == "accept") + { + if (requirements_) + fail_requirements (); + + throw parsing (*name_, + t.line, + t.column, + "accept clause should follow prefer clause"); + } + else + bad_token (what + " alternative clause"); + } + + expect_token (type::rcbrace); + + // Disallow comments. + // + lexer_->comments = false; + + next (t, tt); + } + } + + return r; + } + + dependency_alternatives_parser::token_type dependency_alternatives_parser:: + next (token& t, token_type& tt) + { + t = lexer_->next (); + tt = t.type; + return tt; + } + + dependency_alternatives_parser::token_type dependency_alternatives_parser:: + next_eval (token& t, token_type& tt) + { + t = lexer_->next_eval (); + tt = t.type; + return tt; + } + + dependency_alternatives_parser::token_type dependency_alternatives_parser:: + next_line (token& t, token_type& tt) + { + t = lexer_->next_line ('|'); + tt = t.type; + return tt; + } + + dependency_alternatives_parser::token_type dependency_alternatives_parser:: + next_block (token& t, token_type& tt) + { + t = lexer_->next_block (); + tt = t.type; + return tt; + } + + dependency_alternatives:: + dependency_alternatives (const std::string& s, + const package_name& dependent, + const std::string& name, + uint64_t line, + uint64_t column) + { + using std::string; + + auto vc (parser::split_comment (s)); + + comment = move (vc.second); + + const string& v (vc.first); + buildtime = (v[0] == '*'); + + string::const_iterator b (v.begin ()); + string::const_iterator e (v.end ()); + + if (buildtime) + { + string::size_type p (v.find_first_not_of (spaces, 1)); + b = p == string::npos ? e : b + p; + } + + dependency_alternatives_parser p; + istringstream is (b == v.begin () ? v : string (b, e)); + p.parse (dependent, is, name, line, column, *this); + } + + string dependency_alternatives:: + string () const + { + std::string r (buildtime ? "* " : ""); + + const dependency_alternative* prev (nullptr); + for (const dependency_alternative& da: *this) + { + if (prev != nullptr) + { + r += prev->single_line () ? " |" : "\n|"; + r += !da.single_line () || !prev->single_line () ? '\n' : ' '; + } + + r += da.string (); + prev = &da; + } + + return serializer::merge_comment (r, comment); + } + + // requirement_alternative + // + string requirement_alternative:: + string () const { - if (as.conditional) - o << '?'; + using std::string; + + string r (size () > 1 ? "{" : ""); + + bool first (true); + for (const string& rq: *this) + { + if (!first) + r += ' '; + else + first = false; + + r += rq; + } + + if (size () > 1) + r += '}'; + + if (single_line ()) + { + if (enable) + { + if (!simple ()) + { + r += " ? ("; + r += *enable; + r += ')'; + } + else + { + // Note that the (single) requirement id may or may not be empty. + // + if (!r.empty ()) + r += ' '; + + r += '?'; + + if (!enable->empty ()) + { + r += " ("; + r += *enable; + r += ')'; + } + } + } + + if (reflect) + { + r += ' '; + r += *reflect; + } + } + else + { + r += "\n{"; - if (as.buildtime) - o << '*'; + if (enable) + { + r += "\n enable ("; + r += *enable; + r += ')'; + } - if (as.conditional || as.buildtime) - o << ' '; + if (reflect) + { + if (enable) + r += '\n'; - bool f (true); - for (const dependency& a: as) - o << (f ? (f = false, "") : " | ") << a; + r += "\n reflect\n {\n"; + r += *reflect; + r += " }"; + } - if (!as.comment.empty ()) - o << "; " << as.comment; + r += "\n}"; + } - return o; + return r; + } + + // requirement_alternatives + // + requirement_alternatives:: + requirement_alternatives (const std::string& s, + const package_name& dependent, + const std::string& name, + uint64_t line, + uint64_t column) + { + using std::string; + + auto vc (parser::split_comment (s)); + + comment = move (vc.second); + + const string& v (vc.first); + buildtime = (v[0] == '*'); + + string::const_iterator b (v.begin ()); + string::const_iterator e (v.end ()); + + if (buildtime) + { + string::size_type p (v.find_first_not_of (spaces, 1)); + b = p == string::npos ? e : b + p; + } + + // We will use the dependency alternatives parser to parse the + // representation into a temporary dependency alternatives in the + // requirements mode. Then we will move the dependency alternatives into + // the requirement alternatives using the string representation of the + // dependencies. + // + dependency_alternatives_parser p (true /* requirements */); + istringstream is (b == v.begin () ? v : string (b, e)); + + dependency_alternatives das; + p.parse (dependent, is, name, line, column, das); + + for (dependency_alternative& da: das) + { + requirement_alternative ra (move (da.enable), move (da.reflect)); + + // Also handle the simple requirement. + // + for (dependency& d: da) + ra.push_back (!d.name.empty () ? d.string () : string ()); + + push_back (move (ra)); + } + + // Make sure that the simple requirement is accompanied with a non-empty + // comment. + // + if (simple () && comment.empty ()) + { + // Let's describe the following error cases differently: + // + // requires: ? + // requires: + // + throw parsing (name, + line, + column, + (back ().enable + ? "no comment specified for simple requirement" + : "requirement or comment expected")); + } + } + + std::string requirement_alternatives:: + string () const + { + using std::string; + + string r (buildtime ? "* " : ""); + + const requirement_alternative* prev (nullptr); + for (const requirement_alternative& ra: *this) + { + if (prev != nullptr) + { + r += prev->single_line () ? " |" : "\n|"; + r += !ra.single_line () || !prev->single_line () ? '\n' : ' '; + } + + r += ra.string (); + prev = &ra; + } + + // For better readability separate the comment from the question mark for + // the simple requirement with an empty condition. + // + if (simple () && conditional () && back ().enable->empty ()) + r += ' '; + + return serializer::merge_comment (r, comment); } // build_class_term @@ -1062,7 +2718,7 @@ namespace bpkg } build_class_term:: - build_class_term (build_class_term&& t) + build_class_term (build_class_term&& t) noexcept : operation (t.operation), inverted (t.inverted), simple (t.simple) @@ -1086,13 +2742,13 @@ namespace bpkg } build_class_term& build_class_term:: - operator= (build_class_term&& t) + operator= (build_class_term&& t) noexcept { if (this != &t) { this->~build_class_term (); - // Assume noexcept move-construction. + // Rely on noexcept move-construction. // new (this) build_class_term (move (t)); } @@ -1118,13 +2774,13 @@ namespace bpkg if (!(alnum (c) || c == '_')) throw invalid_argument ( - "class name '" + s + "' starts with '" + c + "'"); + "class name '" + s + "' starts with '" + c + '\''); for (; i != s.size (); ++i) { if (!(alnum (c = s[i]) || c == '+' || c == '-' || c == '_' || c == '.')) throw invalid_argument ( - "class name '" + s + "' contains '" + c + "'"); + "class name '" + s + "' contains '" + c + '\''); } return s[0] == '_'; @@ -1413,102 +3069,42 @@ namespace bpkg match_classes (cs, im, expr, r); } - // text_type + // build_auxiliary // - string - to_string (text_type t) - { - switch (t) - { - case text_type::plain: return "text/plain"; - case text_type::github_mark: return "text/markdown;variant=GFM"; - case text_type::common_mark: return "text/markdown;variant=CommonMark"; - } - - assert (false); // Can't be here. - return string (); - } - - optional<text_type> - to_text_type (const string& t) + optional<pair<string, string>> build_auxiliary:: + parse_value_name (const string& n) { - auto bad_type = [] (const string& d) {throw invalid_argument (d);}; - - // Parse the media type representation (see RFC2045 for details) into the - // type/subtype value and the parameter list. Note: we don't support - // parameter quoting and comments for simplicity. + // Check if the value name matches exactly. // - size_t p (t.find (';')); - const string& tp (p != string::npos ? trim (string (t, 0, p)) : t); - - small_vector<pair<string, string>, 1> ps; + if (n == "build-auxiliary") + return make_pair (string (), string ()); - while (p != string::npos) + // Check if this is a *-build-auxiliary name. + // + if (n.size () > 16 && + n.compare (n.size () - 16, 16, "-build-auxiliary") == 0) { - // Extract parameter name. - // - size_t b (p + 1); - p = t.find ('=', b); - - if (p == string::npos) - bad_type ("missing '='"); - - string n (trim (string (t, b, p - b))); - - // Extract parameter value. - // - b = p + 1; - p = t.find (';', b); - - string v (trim (string (t, - b, - p != string::npos ? p - b : string::npos))); - - ps.emplace_back (move (n), move (v)); + return make_pair (string (n, 0, n.size () - 16), string ()); } - // Calculate the resulting text type, failing on unrecognized media type, - // unexpected parameter name or value. - // - // Note that type, subtype, and parameter names are matched - // case-insensitively. + // Check if this is a build-auxiliary-* name. // - optional<text_type> r; + if (n.size () > 16 && n.compare (0, 16, "build-auxiliary-") == 0) + return make_pair (string (), string (n, 16)); - // Currently only the plain and markdown text types are allowed. Later we - // can potentially introduce some other text types. + // Check if this is a *-build-auxiliary-* name. // - if (icasecmp (tp, "text/plain") == 0) - { - // Currently, we don't expect parameters for plain text. Later we can - // potentially introduce some plain text variants. - // - if (ps.empty ()) - r = text_type::plain; - } - else if (icasecmp (tp, "text/markdown") == 0) + size_t p (n.find ("-build-auxiliary-")); + + if (p != string::npos && + p != 0 && // Not '-build-auxiliary-*'? + p + 17 != n.size () && // Not '*-build-auxiliary-'? + n.find ("-build-auxiliary-", p + 17) == string::npos) // Unambiguous? { - // Currently, a single optional variant parameter with the two possible - // values is allowed for markdown. Later we can potentially introduce - // some other markdown variants. - // - if (ps.empty () || - (ps.size () == 1 && icasecmp (ps[0].first, "variant") == 0)) - { - // Note that markdown variants are matched case-insensitively (see - // RFC7763 for details). - // - string v; - if (ps.empty () || icasecmp (v = move (ps[0].second), "GFM") == 0) - r = text_type::github_mark; - else if (icasecmp (v, "CommonMark") == 0) - r = text_type::common_mark; - } + return make_pair (string (n, 0, p), string (n, p + 17)); } - else if (icasecmp (tp, "text/", 5) != 0) - bad_type ("text type expected"); - return r; + return nullopt; } // test_dependency_type @@ -1533,7 +3129,135 @@ namespace bpkg if (t == "tests") return test_dependency_type::tests; else if (t == "examples") return test_dependency_type::examples; else if (t == "benchmarks") return test_dependency_type::benchmarks; - else throw invalid_argument ("invalid test dependency type '" + t + "'"); + else throw invalid_argument ("invalid test dependency type '" + t + '\''); + } + + + // test_dependency + // + test_dependency:: + test_dependency (std::string v, test_dependency_type t) + : type (t) + { + using std::string; + + // We will use the dependency alternatives parser to parse the + // `<name> [<version-constraint>] ['?' <enable-condition>] [<reflect-config>]` + // representation into a temporary dependency alternatives object. Then we + // will verify that the result has no multiple alternatives/dependency + // packages and unexpected clauses and will move the required information + // (dependency, reflection, etc) into the being created test dependency + // object. + + // Verify that there is no newline characters to forbid the multi-line + // dependency alternatives representation. + // + if (v.find ('\n') != string::npos) + throw invalid_argument ("unexpected <newline>"); + + buildtime = (v[0] == '*'); + + size_t p (v.find_first_not_of (spaces, buildtime ? 1 : 0)); + + if (p == string::npos) + throw invalid_argument ("no package name specified"); + + string::const_iterator b (v.begin () + p); + string::const_iterator e (v.end ()); + + // Extract the dependency package name in advance, to pass it to the + // parser which will use it to verify the reflection variable name. + // + // Note that multiple packages can only be specified in {} to be accepted + // by the parser. In our case such '{' would be interpreted as a part of + // the package name and so would fail complaining about an invalid + // character. Let's handle this case manually to avoid the potentially + // confusing error description. + // + assert (b != e); // We would fail earlier otherwise. + + if (*b == '{') + throw invalid_argument ("only single package allowed"); + + package_name dn; + + try + { + p = v.find_first_of (" \t=<>[(~^", p); // End of the package name. + dn = package_name (string (b, p == string::npos ? e : v.begin () + p)); + } + catch (const invalid_argument& e) + { + throw invalid_argument (string ("invalid package name: ") + e.what ()); + } + + // Parse the value into the temporary dependency alternatives object. + // + dependency_alternatives das; + + try + { + dependency_alternatives_parser p; + istringstream is (b == v.begin () ? v : string (b, e)); + p.parse (dn, is, "" /* name */, 1, 1, das); + } + catch (const manifest_parsing& e) + { + throw invalid_argument (e.description); + } + + // Verify that there are no multiple dependency alternatives. + // + assert (!das.empty ()); // Enforced by the parser. + + if (das.size () != 1) + throw invalid_argument ("unexpected '|'"); + + dependency_alternative& da (das[0]); + + // Verify that there are no multiple dependencies in the alternative. + // + // The parser can never end up with no dependencies in an alternative and + // we already verified that there can't be multiple of them (see above). + // + assert (da.size () == 1); + + // Verify that there are no unexpected clauses. + // + // Note that the require, prefer, and accept clauses can only be present + // in the multi-line representation and we have already verified that this + // is not the case. So there is nothing to verify here. + + // Move the dependency and the enable and reflect clauses into the being + // created test dependency object. + // + static_cast<dependency&> (*this) = move (da[0]); + + enable = move (da.enable); + reflect = move (da.reflect); + } + + string test_dependency:: + string () const + { + std::string r (buildtime + ? "* " + dependency::string () + : dependency::string ()); + + if (enable) + { + r += " ? ("; + r += *enable; + r += ')'; + } + + if (reflect) + { + r += ' '; + r += *reflect; + } + + return r; } // pkg_package_manifest @@ -1590,7 +3314,7 @@ namespace bpkg { throw !source_name.empty () ? parsing (source_name, nv.value_line, nv.value_column, d) - : parsing (d + " in '" + v + "'"); + : parsing (d + " in '" + v + '\''); }; size_t p (v.find ('/')); @@ -1632,42 +3356,115 @@ namespace bpkg return email (move (v), move (c)); } + // Parse the [*-]build-auxiliary[-*] manifest value. + // + // Note that the environment name is expected to already be retrieved using + // build_auxiliary::parse_value_name(). + // + static build_auxiliary + parse_build_auxiliary (const name_value& nv, + string&& env_name, + const string& source_name) + { + auto bad_value = [&nv, &source_name] (const string& d) + { + throw !source_name.empty () + ? parsing (source_name, nv.value_line, nv.value_column, d) + : parsing (d); + }; + + pair<string, string> vc (parser::split_comment (nv.value)); + string& v (vc.first); + string& c (vc.second); + + if (v.empty ()) + bad_value ("empty build auxiliary configuration name pattern"); + + return build_auxiliary (move (env_name), move (v), move (c)); + } + + // Parse the [*-]build-bot manifest value and append it to the specified + // custom bot public keys list. Make sure the specified key is not empty and + // is not a duplicate and throw parsing if that's not the case. + // + // Note: value name is not used by this function (and so can be moved out, + // etc before the call). + // + static void + parse_build_bot (const name_value& nv, const string& source_name, strings& r) + { + const string& v (nv.value); + + auto bad_value = [&nv, &source_name, &v] (const string& d, + bool add_key = true) + { + throw !source_name.empty () + ? parsing (source_name, nv.value_line, nv.value_column, d) + : parsing (!add_key ? d : (d + ":\n" + v)); + }; + + if (v.empty ()) + bad_value ("empty custom build bot public key", false /* add_key */); + + if (find (r.begin (), r.end (), v) != r.end ()) + bad_value ("duplicate custom build bot public key"); + + r.push_back (v); + } + const version stub_version (0, "0", nullopt, nullopt, 0); + // Parse until next() returns end-of-manifest value. + // static void parse_package_manifest ( - parser& p, - name_value nv, - const function<package_manifest::translate_function>& tf, + const string& name, + const function<name_value ()>& next, + const function<package_manifest::translate_function>& translate, bool iu, - bool cd, + bool cv, package_manifest_flags fl, package_manifest& m) { - auto bad_name ([&p, &nv](const string& d) { - throw parsing (p.name (), nv.name_line, nv.name_column, d);}); + name_value nv; - auto bad_value ([&p, &nv](const string& d) { - throw parsing (p.name (), nv.value_line, nv.value_column, d);}); - - // Make sure this is the start and we support the version. - // - if (!nv.name.empty ()) - bad_name ("start of package manifest expected"); + auto bad_name ([&name, &nv](const string& d) { + throw parsing (name, nv.name_line, nv.name_column, d);}); - if (nv.value != "1") - bad_value ("unsupported format version"); + auto bad_value ([&name, &nv](const string& d) { + throw parsing (name, nv.value_line, nv.value_column, d);}); - auto parse_email = [&bad_name] (const name_value& nv, - optional<email>& r, - const char* what, - const string& source_name, - bool empty = false) + auto parse_email = [&bad_name, &name] (const name_value& nv, + optional<email>& r, + const char* what, + bool empty = false) { if (r) bad_name (what + string (" email redefinition")); - r = bpkg::parse_email (nv, what, source_name, empty); + r = bpkg::parse_email (nv, what, name, empty); + }; + + // Parse the [*-]build-auxiliary[-*] manifest value and append it to the + // specified build auxiliary list. Make sure that the list contains not + // more than one entry with unspecified environment name and throw parsing + // if that's not the case. Also make sure that there are no entry + // redefinitions (multiple entries with the same environment name). + // + auto parse_build_auxiliary = [&bad_name, &name] (const name_value& nv, + string&& en, + vector<build_auxiliary>& r) + { + build_auxiliary a (bpkg::parse_build_auxiliary (nv, move (en), name)); + + if (find_if (r.begin (), r.end (), + [&a] (const build_auxiliary& ba) + { + return ba.environment_name == a.environment_name; + }) != r.end ()) + bad_name ("build auxiliary environment redefinition"); + + r.push_back (move (a)); }; auto parse_url = [&bad_value] (const string& v, @@ -1731,11 +3528,190 @@ namespace bpkg } }; + // Note: the n argument is the distribution name length. + // + auto parse_distribution = [&bad_name, &bad_value] (string&& nm, size_t n, + string&& vl) + { + size_t p (nm.find ('-')); + + // Distribution-related manifest value name always has a dash-starting + // suffix (-name, etc). + // + assert (p != string::npos); + + if (p < n) + bad_name ("distribution name '" + string (nm, 0, n) + "' contains '-'"); + + if (vl.empty ()) + bad_value ("empty package distribution value"); + + return distribution_name_value (move (nm), move (vl)); + }; + + auto add_distribution = [&m, &bad_name] (distribution_name_value&& nv, + bool unique) + { + vector<distribution_name_value>& dvs (m.distribution_values); + + if (unique && + find_if (dvs.begin (), dvs.end (), + [&nv] (const distribution_name_value& dnv) + {return dnv.name == nv.name;}) != dvs.end ()) + { + bad_name ("package distribution value redefinition"); + } + + dvs.push_back (move (nv)); + }; + auto flag = [fl] (package_manifest_flags f) { return (fl & f) != package_manifest_flags::none; }; + // Based on the buildfile path specified via the `*-build[2]` value name + // or the `build-file` value set the manifest's alt_naming flag if absent + // and verify that it doesn't change otherwise. If it does, then return + // the error description and nullopt otherwise. + // + auto alt_naming = [&m] (const string& p) -> optional<string> + { + assert (!p.empty ()); + + bool an (p.back () == '2'); + + if (!m.alt_naming) + m.alt_naming = an; + else if (*m.alt_naming != an) + return string (*m.alt_naming ? "alternative" : "standard") + + " buildfile naming scheme is already used"; + + return nullopt; + }; + + // Try to parse and verify the buildfile path specified via the + // `*-build[2]` value name or the `build-file` value and set the + // manifest's alt_naming flag. On success return the normalized path with + // the suffix stripped and nullopt and the error description + // otherwise. Expects that the prefix is not empty. + // + // Specifically, verify that the path doesn't contain backslashes, is + // relative, doesn't refer outside the packages's build subdirectory, and + // was not specified yet. Also verify that the file name is not empty. + // + auto parse_buildfile_path = + [&m, &alt_naming] (string&& p, string& err) -> optional<path> + { + if (optional<string> e = alt_naming (p)) + { + err = move (*e); + return nullopt; + } + + // Verify that the path doesn't contain backslashes which would be + // interpreted differently on Windows and POSIX. + // + if (p.find ('\\') != string::npos) + { + err = "backslash in package buildfile path"; + return nullopt; + } + + // Strip the '(-|.)build' suffix. + // + size_t n (*m.alt_naming ? 7 : 6); + assert (p.size () > n); + + p.resize (p.size () - n); + + try + { + path f (move (p)); + + // Fail if the value name is something like `config/-build`. + // + if (f.to_directory ()) + { + err = "empty package buildfile name"; + return nullopt; + } + + if (f.absolute ()) + { + err = "absolute package buildfile path"; + return nullopt; + } + + // Verify that the path refers inside the package's build/ + // subdirectory. + // + f.normalize (); // Note: can't throw since the path is relative. + + if (dir_path::traits_type::parent (*f.begin ())) + { + err = "package buildfile path refers outside build/ subdirectory"; + return nullopt; + } + + // Check for duplicates. + // + const vector<buildfile>& bs (m.buildfiles); + const vector<path>& bps (m.buildfile_paths); + + if (find_if (bs.begin (), bs.end (), + [&f] (const auto& v) {return v.path == f;}) + != bs.end () || + find (bps.begin (), bps.end (), f) != bps.end ()) + { + err = "package buildfile redefinition"; + return nullopt; + } + + return f; + } + catch (const invalid_path&) + { + err = "invalid package buildfile path"; + return nullopt; + } + }; + + // Return the package build configuration with the specified name, if + // already exists. If no configuration matches, then create one, if + // requested, and throw manifest_parsing otherwise. If the new + // configuration creation is not allowed, then the description for a + // potential manifest_parsing exception needs to also be specified. + // + auto build_conf = [&m, &bad_name] (string&& nm, + bool create = true, + const string& desc = "") + -> build_package_config& + { + // The error description must only be specified if the creation of the + // package build configuration is not allowed. + // + assert (desc.empty () == create); + + small_vector<build_package_config, 1>& cs (m.build_configs); + + auto i (find_if (cs.begin (), cs.end (), + [&nm] (const build_package_config& c) + {return c.name == nm;})); + + if (i != cs.end ()) + return *i; + + if (!create) + bad_name (desc + ": no build package configuration '" + nm + '\''); + + // Add the new build configuration (arguments, builds, etc will come + // later). + // + cs.emplace_back (move (nm)); + return cs.back (); + }; + // Cache the upstream version manifest value and validate whether it's // allowed later, after the version value is parsed. // @@ -1743,18 +3719,36 @@ namespace bpkg // We will cache the depends and the test dependency manifest values to // parse and, if requested, complete the version constraints later, after - // the version value is parsed. + // the version value is parsed. We will also cache the requires values to + // parse them later, after the package name is parsed. // vector<name_value> dependencies; + vector<name_value> requirements; small_vector<name_value, 1> tests; - // We will cache the description and its type values to validate them - // later, after both are parsed. + // We will cache the descriptions and changes and their type values to + // validate them later, after all are parsed. // optional<name_value> description; optional<name_value> description_type; + optional<name_value> package_description; + optional<name_value> package_description_type; + vector<name_value> changes; + optional<name_value> changes_type; + + // It doesn't make sense for only emails to be specified for a package + // build configuration. Thus, we will cache the build configuration email + // manifest values to parse them later, after all other build + // configuration values are parsed, and to make sure that the build + // configurations they refer to are also specified. + // + vector<name_value> build_config_emails; + vector<name_value> build_config_warning_emails; + vector<name_value> build_config_error_emails; - for (nv = p.next (); !nv.empty (); nv = p.next ()) + m.build_configs.emplace_back ("default"); + + for (nv = next (); !nv.empty (); nv = next ()) { string& n (nv.name); string& v (nv.value); @@ -1793,9 +3787,9 @@ namespace bpkg if (m.version.release && m.version.release->empty ()) bad_value ("invalid package version release"); - if (tf) + if (translate) { - tf (m.version); + translate (m.version); // Re-validate the version after the translation. // @@ -1821,6 +3815,55 @@ namespace bpkg upstream_version = move (nv); } + else if (n == "type") + { + if (m.type) + bad_name ("package type redefinition"); + + if (v.empty () || v.find (',') == 0) + bad_value ("empty package type"); + + m.type = move (v); + } + else if (n == "language") + { + // Strip the language extra information, if present. + // + size_t p (v.find (',')); + if (p != string::npos) + v.resize (p); + + // Determine the language impl flag. + // + bool impl (false); + p = v.find ('='); + if (p != string::npos) + { + string s (trim (string (v, p + 1))); + if (s != "impl") + bad_value (!s.empty () + ? "unexpected '" + s + "' value after '='" + : "expected 'impl' after '='"); + + impl = true; + + v.resize (p); + } + + // Finally, validate and add the language. + // + trim_right (v); + + if (v.empty ()) + bad_value ("empty package language"); + + if (find_if (m.languages.begin (), m.languages.end (), + [&v] (const language& l) {return l.name == v;}) != + m.languages.end ()) + bad_value ("duplicate package language"); + + m.languages.emplace_back (move (v), impl); + } else if (n == "project") { if (m.project) @@ -1877,28 +3920,28 @@ namespace bpkg if (description) { if (description->name == "description-file") - bad_name ("package description and description-file are " + bad_name ("project description and description file are " "mutually exclusive"); else - bad_name ("package description redefinition"); + bad_name ("project description redefinition"); } if (v.empty ()) - bad_value ("empty package description"); + bad_value ("empty project description"); description = move (nv); } else if (n == "description-file") { if (flag (package_manifest_flags::forbid_file)) - bad_name ("package description-file not allowed"); + bad_name ("project description file not allowed"); if (description) { if (description->name == "description-file") - bad_name ("package description-file redefinition"); + bad_name ("project description file redefinition"); else - bad_name ("package description-file and description are " + bad_name ("project description file and description are " "mutually exclusive"); } @@ -1907,32 +3950,69 @@ namespace bpkg else if (n == "description-type") { if (description_type) - bad_name ("package description-type redefinition"); + bad_name ("project description type redefinition"); description_type = move (nv); } + else if (n == "package-description") + { + if (package_description) + { + if (package_description->name == "package-description-file") + bad_name ("package description and description file are " + "mutually exclusive"); + else + bad_name ("package description redefinition"); + } + + if (v.empty ()) + bad_value ("empty package description"); + + package_description = move (nv); + } + else if (n == "package-description-file") + { + if (flag (package_manifest_flags::forbid_file)) + bad_name ("package description file not allowed"); + + if (package_description) + { + if (package_description->name == "package-description-file") + bad_name ("package description file redefinition"); + else + bad_name ("package description file and description are " + "mutually exclusive"); + } + + package_description = move (nv); + } + else if (n == "package-description-type") + { + if (package_description_type) + bad_name ("package description type redefinition"); + + package_description_type = move (nv); + } else if (n == "changes") { if (v.empty ()) bad_value ("empty package changes specification"); - m.changes.emplace_back (move (v)); + changes.emplace_back (move (nv)); } else if (n == "changes-file") { if (flag (package_manifest_flags::forbid_file)) bad_name ("package changes-file not allowed"); - auto vc (parser::split_comment (v)); - path p (move (vc.first)); - - if (p.empty ()) - bad_value ("no path in package changes-file"); - - if (p.absolute ()) - bad_value ("package changes-file path is absolute"); + changes.emplace_back (move (nv)); + } + else if (n == "changes-type") + { + if (changes_type) + bad_name ("package changes type redefinition"); - m.changes.emplace_back (move (p), move (vc.second)); + changes_type = move (nv); } else if (n == "url") { @@ -1943,7 +4023,7 @@ namespace bpkg } else if (n == "email") { - parse_email (nv, m.email, "project", p.name ()); + parse_email (nv, m.email, "project"); } else if (n == "doc-url") { @@ -1968,19 +4048,19 @@ namespace bpkg } else if (n == "package-email") { - parse_email (nv, m.package_email, "package", p.name ()); + parse_email (nv, m.package_email, "package"); } else if (n == "build-email") { - parse_email (nv, m.build_email, "build", p.name (), true /* empty */); + parse_email (nv, m.build_email, "build", true /* empty */); } else if (n == "build-warning-email") { - parse_email (nv, m.build_warning_email, "build warning", p.name ()); + parse_email (nv, m.build_warning_email, "build warning"); } else if (n == "build-error-email") { - parse_email (nv, m.build_error_email, "build error", p.name ()); + parse_email (nv, m.build_error_email, "build error"); } else if (n == "priority") { @@ -2031,60 +4111,222 @@ namespace bpkg m.license_alternatives.push_back (move (l)); } + else if (n == "depends") + { + dependencies.push_back (move (nv)); + } else if (n == "requires") { - // Allow specifying ?* in any order. - // - size_t n (v.size ()); - size_t cond ((n > 0 && v[0] == '?') || (n > 1 && v[1] == '?') ? 1 : 0); - size_t btim ((n > 0 && v[0] == '*') || (n > 1 && v[1] == '*') ? 1 : 0); - + requirements.push_back (move (nv)); + } + else if (n == "builds") + { + m.builds.push_back ( + parse_build_class_expr (nv, m.builds.empty (), name)); + } + else if (n == "build-include") + { + m.build_constraints.push_back ( + parse_build_constraint (nv, false /* exclusion */, name)); + } + else if (n == "build-exclude") + { + m.build_constraints.push_back ( + parse_build_constraint (nv, true /* exclusion */, name)); + } + else if (optional<pair<string, string>> ba = + build_auxiliary::parse_value_name (n)) + { + if (ba->first.empty ()) // build-auxiliary*? + { + parse_build_auxiliary (nv, move (ba->second), m.build_auxiliaries); + } + else // *-build-auxiliary* + { + build_package_config& bc (build_conf (move (ba->first))); + parse_build_auxiliary (nv, move (ba->second), bc.auxiliaries); + } + } + else if (n == "build-bot") + { + parse_build_bot (nv, name, m.build_bot_keys); + } + else if (n.size () > 13 && + n.compare (n.size () - 13, 13, "-build-config") == 0) + { auto vc (parser::split_comment (v)); - const string& vl (vc.first); - requirement_alternatives ra (cond != 0, btim != 0, move (vc.second)); + n.resize (n.size () - 13); - string::const_iterator b (vl.begin ()); - string::const_iterator e (vl.end ()); + build_package_config& bc (build_conf (move (n))); - if (ra.conditional || ra.buildtime) - { - string::size_type p (vl.find_first_not_of (spaces, cond + btim)); - b = p == string::npos ? e : b + p; - } + if (!bc.arguments.empty () || !bc.comment.empty ()) + bad_name ("build configuration redefinition"); - list_parser lp (b, e, '|'); - for (string lv (lp.next ()); !lv.empty (); lv = lp.next ()) - ra.push_back (lv); + bc.arguments = move (vc.first); + bc.comment = move (vc.second); + } + else if (n.size () > 7 && n.compare (n.size () - 7, 7, "-builds") == 0) + { + n.resize (n.size () - 7); - if (ra.empty () && ra.comment.empty ()) - bad_value ("empty package requirement specification"); + build_package_config& bc (build_conf (move (n))); - m.requirements.push_back (move (ra)); + bc.builds.push_back ( + parse_build_class_expr (nv, bc.builds.empty (), name)); } - else if (n == "builds") + else if (n.size () > 14 && + n.compare (n.size () - 14, 14, "-build-include") == 0) { - m.builds.push_back ( - parse_build_class_expr (nv, m.builds.empty (), p.name ())); + n.resize (n.size () - 14); + + build_package_config& bc (build_conf (move (n))); + + bc.constraints.push_back ( + parse_build_constraint (nv, false /* exclusion */, name)); } - else if (n == "build-include") + else if (n.size () > 14 && + n.compare (n.size () - 14, 14, "-build-exclude") == 0) { - m.build_constraints.push_back ( - parse_build_constraint (nv, false /* exclusion */, p.name ())); + n.resize (n.size () - 14); + + build_package_config& bc (build_conf (move (n))); + + bc.constraints.push_back ( + parse_build_constraint (nv, true /* exclusion */, name)); } - else if (n == "build-exclude") + else if (n.size () > 10 && + n.compare (n.size () - 10, 10, "-build-bot") == 0) { - m.build_constraints.push_back ( - parse_build_constraint (nv, true /* exclusion */, p.name ())); + n.resize (n.size () - 10); + + build_package_config& bc (build_conf (move (n))); + parse_build_bot (nv, name, bc.bot_keys); } - else if (n == "depends") + else if (n.size () > 12 && + n.compare (n.size () - 12, 12, "-build-email") == 0) { - dependencies.push_back (move (nv)); + n.resize (n.size () - 12); + build_config_emails.push_back (move (nv)); + } + else if (n.size () > 20 && + n.compare (n.size () - 20, 20, "-build-warning-email") == 0) + { + n.resize (n.size () - 20); + build_config_warning_emails.push_back (move (nv)); } - else if (n == "tests" || n == "examples" || n == "benchmarks") + else if (n.size () > 18 && + n.compare (n.size () - 18, 18, "-build-error-email") == 0) { + n.resize (n.size () - 18); + build_config_error_emails.push_back (move (nv)); + } + // @@ TMP time to drop *-0.14.0? + // + else if (n == "tests" || n == "tests-0.14.0" || + n == "examples" || n == "examples-0.14.0" || + n == "benchmarks" || n == "benchmarks-0.14.0") + { + // Strip the '-0.14.0' suffix from the value name, if present. + // + size_t p (n.find ('-')); + if (p != string::npos) + n.resize (p); + tests.push_back (move (nv)); } + else if (n == "bootstrap-build" || n == "bootstrap-build2") + { + if (optional<string> e = alt_naming (n)) + bad_name (*e); + + if (m.bootstrap_build) + bad_name ("package " + n + " redefinition"); + + m.bootstrap_build = move (v); + } + else if (n == "root-build" || n == "root-build2") + { + if (optional<string> e = alt_naming (n)) + bad_name (*e); + + if (m.root_build) + bad_name ("package " + n + " redefinition"); + + m.root_build = move (v); + } + else if ((n.size () > 6 && n.compare (n.size () - 6, 6, "-build") == 0) || + (n.size () > 7 && n.compare (n.size () - 7, 7, "-build2") == 0)) + { + string err; + if (optional<path> p = parse_buildfile_path (move (n), err)) + m.buildfiles.push_back (buildfile (move (*p), move (v))); + else + bad_name (err); + } + else if (n == "build-file") + { + if (flag (package_manifest_flags::forbid_file)) + bad_name ("package build-file not allowed"); + + // Verify that the buildfile extension is either build or build2. + // + if ((v.size () > 6 && v.compare (v.size () - 6, 6, ".build") == 0) || + (v.size () > 7 && v.compare (v.size () - 7, 7, ".build2") == 0)) + { + string err; + if (optional<path> p = parse_buildfile_path (move (v), err)) + { + // Verify that the resulting path differs from bootstrap and root. + // + const string& s (p->string ()); + if (s == "bootstrap" || s == "root") + bad_value (s + " not allowed"); + + m.buildfile_paths.push_back (move (*p)); + } + else + bad_value (err); + } + else + bad_value ("path with build or build2 extension expected"); + + } + else if (n.size () > 5 && n.compare (n.size () - 5, 5, "-name") == 0) + { + add_distribution ( + parse_distribution (move (n), n.size () - 5, move (v)), + false /* unique */); + } + // Note: must precede the check for the "-version" suffix. + // + else if (n.size () > 22 && + n.compare (n.size () - 22, 22, "-to-downstream-version") == 0) + { + add_distribution ( + parse_distribution (move (n), n.size () - 22, move (v)), + false /* unique */); + } + // Note: must follow the check for "upstream-version". + // + else if (n.size () > 8 && n.compare (n.size () - 8, 8, "-version") == 0) + { + // If the value is forbidden then throw, but only after the name is + // validated. Thus, check for that before we move the value from. + // + bool bad (v == "$" && + flag (package_manifest_flags::forbid_incomplete_values)); + + // Can throw. + // + distribution_name_value d ( + parse_distribution (move (n), n.size () - 8, move (v))); + + if (bad) + bad_value ("$ not allowed"); + + add_distribution (move (d), true /* unique */); + } else if (n == "location") { if (flag (package_manifest_flags::forbid_location)) @@ -2165,23 +4407,24 @@ namespace bpkg m.upstream_version = move (nv.value); } - // Verify that description is specified if the description type is - // specified. - // - if (description_type && !description) - bad_value ("no package description for specified description type"); - - // Validate (and set) description and its type. + // Parse and validate a text/file manifest value and its respective type + // value, if present. Return a typed_text_file object. // - if (description) + auto parse_text_file = [iu, &nv, &bad_value] (name_value&& text_file, + optional<name_value>&& type, + const char* what) + -> typed_text_file { + typed_text_file r; + // Restore as bad_value() uses its line/column. // - nv = move (*description); + nv = move (text_file); string& v (nv.value); + const string& n (nv.name); - if (nv.name == "description-file") + if (n.size () > 5 && n.compare (n.size () - 5, 5, "-file") == 0) { auto vc (parser::split_comment (v)); @@ -2192,112 +4435,223 @@ namespace bpkg } catch (const invalid_path& e) { - bad_value (string ("invalid package description file: ") + - e.what ()); + bad_value (string ("invalid ") + what + " file: " + e.what ()); } if (p.empty ()) - bad_value ("no path in package description-file"); + bad_value (string ("no path in ") + what + " file"); if (p.absolute ()) - bad_value ("package description-file path is absolute"); + bad_value (string (what) + " file path is absolute"); - m.description = text_file (move (p), move (vc.second)); + r = typed_text_file (move (p), move (vc.second)); } else - m.description = text_file (move (v)); + r = typed_text_file (move (v)); - if (description_type) - m.description_type = move (description_type->value); + if (type) + r.type = move (type->value); - // Verify the description type. + // Verify the text type. // try { - m.effective_description_type (iu); + r.effective_type (iu); } catch (const invalid_argument& e) { - if (description_type) + if (type) { - // Restore as bad_value() uses its line/column. + // Restore as bad_value() uses its line/column. Note that we don't + // need to restore the moved out type value. // - nv = move (*description_type); + nv = move (*type); - bad_value (string ("invalid package description type: ") + - e.what ()); + bad_value (string ("invalid ") + what + " type: " + e.what ()); } else - bad_value (string ("invalid package description file: ") + - e.what ()); + { + // Note that this can only happen due to inability to guess the + // type from the file extension. Let's help the user here a bit. + // + assert (r.file); + + bad_value (string ("invalid ") + what + " file: " + e.what () + + " (use " + string (n, 0, n.size () - 5) + + "-type manifest value to specify explicitly)"); + } } - } - // Now, when the version manifest value is parsed, we can parse the - // dependencies and complete their constraints, if requested. + return r; + }; + + // As above but also accepts nullopt as the text_file argument, in which + // case throws manifest_parsing if the type is specified and return + // nullopt otherwise. // - auto parse_dependency = [&m, cd, &flag, &bad_value] (string&& d, - const char* what) + auto parse_text_file_opt = [&nv, &bad_name, &parse_text_file] + (optional<name_value>&& text_file, + optional<name_value>&& type, + const char* what) -> optional<typed_text_file> { - using iterator = string::const_iterator; - - iterator b (d.begin ()); - iterator i (b); - iterator ne (b); // End of name. - iterator e (d.end ()); - - // Find end of name (ne). - // - // Grep for '=<>([~^' in the bpkg source code and update, if changed. + // Verify that the text/file value is specified if the type value is + // specified. // - const string cb ("=<>([~^"); - for (char c; i != e && cb.find (c = *i) == string::npos; ++i) + if (!text_file) { - if (!space (c)) - ne = i + 1; + if (type) + { + // Restore as bad_name() uses its line/column. + // + nv = move (*type); + + bad_name (string ("no ") + what + " for specified type"); + } + + return nullopt; } - package_name nm; + return parse_text_file (move (*text_file), move (type), what); + }; - try + // Parse the project/package descriptions/types. + // + m.description = parse_text_file_opt (move (description), + move (description_type), + "project description"); + + m.package_description = + parse_text_file_opt (move (package_description), + move (package_description_type), + "package description"); + + // Parse the package changes/types. + // + // Note: at the end of the loop the changes_type variable may contain + // value in unspecified state but we can still check for the value + // presence. + // + for (name_value& c: changes) + { + // Move the changes_type value from for the last changes entry. + // + m.changes.push_back ( + parse_text_file (move (c), + (&c != &changes.back () + ? optional<name_value> (changes_type) + : move (changes_type)), + "changes")); + } + + // If there are multiple changes and the changes type is not explicitly + // specified, then verify that all changes effective types are the same. + // Note that in the "ignore unknown" mode there can be unresolved + // effective types which we just skip. + // + if (changes.size () > 1 && !changes_type) + { + optional<text_type> type; + + for (size_t i (0); i != m.changes.size (); ++i) { - nm = package_name (i == e ? move (d) : string (b, ne)); + const typed_text_file& c (m.changes[i]); + + if (optional<text_type> t = c.effective_type (iu)) + { + if (!type) + { + type = *t; + } + else if (*t != *type) + { + // Restore as bad_value() uses its line/column. + // + nv = move (changes[i]); + + bad_value ("changes type '" + to_string (*t) + "' differs from " + + " previous type '" + to_string (*type) + "'"); + } + } } - catch (const invalid_argument& e) + } + + // Parse the build configuration emails. + // + // Note: the argument can only be one of the build_config_*emails + // variables (see above) to distinguish between the email kinds. + // + auto parse_build_config_emails = [&nv, + &build_config_emails, + &build_config_warning_emails, + &build_config_error_emails, + &build_conf, + &parse_email] + (vector<name_value>&& emails) + { + enum email_kind {build, warning, error}; + + email_kind ek ( + &emails == &build_config_emails ? email_kind::build : + &emails == &build_config_warning_emails ? email_kind::warning : + email_kind::error); + + // The argument can only be one of the build_config_*emails variables. + // + assert (ek != email_kind::error || &emails == &build_config_error_emails); + + for (name_value& e: emails) { - bad_value (string ("invalid ") + what + " package name: " + - e.what ()); + // Restore as bad_name() and bad_value() use its line/column. + // + nv = move (e); + + build_package_config& bc ( + build_conf (move (nv.name), + false /* create */, + "stray build notification email")); + + parse_email ( + nv, + (ek == email_kind::build ? bc.email : + ek == email_kind::warning ? bc.warning_email : + bc.error_email), + (ek == email_kind::build ? "build configuration" : + ek == email_kind::warning ? "build configuration warning" : + "build configuration error"), + ek == email_kind::build /* empty */); } + }; - dependency r; + parse_build_config_emails (move (build_config_emails)); + parse_build_config_emails (move (build_config_warning_emails)); + parse_build_config_emails (move (build_config_error_emails)); - if (i == e) - r = dependency {move (nm), nullopt}; - else + // Now, when the version manifest value is parsed, we can parse the + // dependencies and complete their constraints, if requested. + // + auto complete_constraint = [&m, cv, &flag] (auto&& dep) + { + if (dep.constraint) + try { - try - { - version_constraint vc (string (i, e)); - - if (!vc.complete () && - flag (package_manifest_flags::forbid_incomplete_dependencies)) - bad_value ("$ not allowed"); + version_constraint& vc (*dep.constraint); - // Complete the constraint. - // - if (cd) - vc = vc.effective (m.version); + if (!vc.complete () && + flag (package_manifest_flags::forbid_incomplete_values)) + throw invalid_argument ("$ not allowed"); - r = dependency {move (nm), move (vc)}; - } - catch (const invalid_argument& e) - { - bad_value (string ("invalid ") + what + " package constraint: " + - e.what ()); - } + // Complete the constraint. + // + if (cv) + vc = vc.effective (m.version); + } + catch (const invalid_argument& e) + { + throw invalid_argument ("invalid package constraint '" + + dep.constraint->string () + "': " + e.what ()); } - return r; + return move (dep); }; // Parse the regular dependencies. @@ -2306,36 +4660,40 @@ namespace bpkg { nv = move (d); // Restore as bad_value() uses its line/column. - const string& v (nv.value); - - // Allow specifying ?* in any order. + // Parse dependency alternatives. // - size_t n (v.size ()); - size_t cond ((n > 0 && v[0] == '?') || (n > 1 && v[1] == '?') ? 1 : 0); - size_t btim ((n > 0 && v[0] == '*') || (n > 1 && v[1] == '*') ? 1 : 0); - - auto vc (parser::split_comment (v)); - - const string& vl (vc.first); - dependency_alternatives da (cond != 0, btim != 0, move (vc.second)); + try + { + dependency_alternatives das (nv.value, + m.name, + name, + nv.value_line, + nv.value_column); - string::const_iterator b (vl.begin ()); - string::const_iterator e (vl.end ()); + for (dependency_alternative& da: das) + { + for (dependency& d: da) + d = complete_constraint (move (d)); + } - if (da.conditional || da.buildtime) + m.dependencies.push_back (move (das)); + } + catch (const invalid_argument& e) { - string::size_type p (vl.find_first_not_of (spaces, cond + btim)); - b = p == string::npos ? e : b + p; + bad_value (e.what ()); } + } - list_parser lp (b, e, '|'); - for (string lv (lp.next ()); !lv.empty (); lv = lp.next ()) - da.push_back (parse_dependency (move (lv), "prerequisite")); - - if (da.empty ()) - bad_value ("empty package dependency specification"); - - m.dependencies.push_back (da); + // Parse the requirements. + // + for (const name_value& r: requirements) + { + m.requirements.push_back ( + requirement_alternatives (r.value, + m.name, + name, + r.value_line, + r.value_column)); } // Parse the test dependencies. @@ -2344,44 +4702,126 @@ namespace bpkg { nv = move (t); // Restore as bad_value() uses its line/column. - string& v (nv.value); - - bool b (v[0] == '*'); - size_t p (v.find_first_not_of (spaces, b ? 1 : 0)); - - if (p == string::npos) - bad_value ("no " + nv.name + " package name specified"); - - dependency d (parse_dependency (p == 0 ? move (v) : string (v, p), - nv.name.c_str ())); - try { - m.tests.emplace_back ( - move (d.name), - to_test_dependency_type (nv.name), - b, - move (d.constraint)); + m.tests.push_back ( + complete_constraint ( + test_dependency (move (nv.value), + to_test_dependency_type (nv.name)))); } - catch (const invalid_argument&) + catch (const invalid_argument& e) { - // to_test_dependency_type() can't throw since the type string is - // already validated. - // - assert (false); + bad_value (e.what ()); } } - if (m.description && - !m.description_type && - flag (package_manifest_flags::require_description_type)) - bad_name ("no package description type specified"); + // Now, when the version manifest value is parsed, we complete the + // <distribution>-version values, if requested. + // + if (cv) + { + for (distribution_name_value& nv: m.distribution_values) + { + const string& n (nv.name); + string& v (nv.value); + + if (v == "$" && + (n.size () > 8 && n.compare (n.size () - 8, 8, "-version") == 0) && + n.find ('-') == n.size () - 8) + { + v = version (default_epoch (m.version), + move (m.version.upstream), + nullopt /* release */, + nullopt /* revision */, + 0 /* iteration */).string (); + } + } + } if (!m.location && flag (package_manifest_flags::require_location)) bad_name ("no package location specified"); if (!m.sha256sum && flag (package_manifest_flags::require_sha256sum)) bad_name ("no package sha256sum specified"); + + if (flag (package_manifest_flags::require_text_type)) + { + if (m.description && !m.description->type) + bad_name ("no project description type specified"); + + if (m.package_description && !m.package_description->type) + bad_name ("no package description type specified"); + + // Note that changes either all have the same explicitly specified type + // or have no type. + // + if (!m.changes.empty () && !m.changes.front ().type) + { + // @@ TMP To support older repositories allow absent changes type + // until toolchain 0.16.0 is released. + // + // Note that for such repositories the packages may not have + // changes values other than plan text. Thus, we can safely set + // this type, if they are absent, so that the caller can always + // be sure that these values are always present for package + // manifest lists. + //bad_name ("no package changes type specified"); + for (typed_text_file& c: m.changes) + c.type = "text/plain"; + } + } + + if (!m.bootstrap_build && + flag (package_manifest_flags::require_bootstrap_build)) + { + // @@ TMP To support older repositories allow absent bootstrap build + // and alt_naming until toolchain 0.15.0 is released. + // + // Note that for such repositories the packages may not have any + // need for the bootstrap buildfile (may not have any dependency + // clauses, etc). Thus, we can safely set the bootstrap build and + // alt_naming values to an empty string and false, respectively, + // if they are absent, so that the caller can always be sure that + // these values are always present for package manifest lists. + // + // Note: don't forget to uncomment no-bootstrap test in + // tests/manifest/testscript when removing this workaround. + // + // bad_name ("no package bootstrap build specified"); + m.bootstrap_build = "project = " + m.name.string () + '\n'; + m.alt_naming = false; + } + } + + static void + parse_package_manifest ( + parser& p, + name_value nv, + const function<package_manifest::translate_function>& tf, + bool iu, + bool cv, + package_manifest_flags fl, + package_manifest& m) + { + // Make sure this is the start and we support the version. + // + if (!nv.name.empty ()) + throw parsing (p.name (), nv.name_line, nv.name_column, + "start of package manifest expected"); + + if (nv.value != "1") + throw parsing (p.name (), nv.value_line, nv.value_column, + "unsupported format version"); + + // Note that we rely on "small function object" optimization here. + // + parse_package_manifest (p.name (), + [&p] () {return p.next ();}, + tf, + iu, + cv, + fl, + m); } package_manifest @@ -2391,12 +4831,13 @@ namespace bpkg p, move (nv), iu, - false /* complete_depends */, + false /* complete_values */, package_manifest_flags::forbid_file | - package_manifest_flags::require_description_type | - package_manifest_flags::require_location | package_manifest_flags::forbid_fragment | - package_manifest_flags::forbid_incomplete_dependencies); + package_manifest_flags::forbid_incomplete_values | + package_manifest_flags::require_location | + package_manifest_flags::require_text_type | + package_manifest_flags::require_bootstrap_build); } // package_manifest @@ -2405,10 +4846,10 @@ namespace bpkg package_manifest (manifest_parser& p, const function<translate_function>& tf, bool iu, - bool cd, + bool cv, package_manifest_flags fl) { - parse_package_manifest (p, p.next (), tf, iu, cd, fl, *this); + parse_package_manifest (p, p.next (), tf, iu, cv, fl, *this); // Make sure this is the end. // @@ -2419,11 +4860,42 @@ namespace bpkg } package_manifest:: - package_manifest (manifest_parser& p, + package_manifest (const string& name, + vector<name_value>&& vs, + const function<translate_function>& tf, + bool iu, + bool cv, + package_manifest_flags fl) + { + auto i (vs.begin ()); + auto e (vs.end ()); + + // Note that we rely on "small function object" optimization here. + // + parse_package_manifest (name, + [&i, &e] () + { + return i != e ? move (*i++) : name_value (); + }, + tf, + iu, + cv, + fl, + *this); + } + + package_manifest:: + package_manifest (const string& name, + vector<name_value>&& vs, bool iu, - bool cd, + bool cv, package_manifest_flags fl) - : package_manifest (p, function<translate_function> (), iu, cd, fl) + : package_manifest (name, + move (vs), + function<translate_function> (), + iu, + cv, + fl) { } @@ -2431,197 +4903,748 @@ namespace bpkg package_manifest (manifest_parser& p, name_value nv, bool iu, - bool cd, + bool cv, package_manifest_flags fl) { parse_package_manifest ( - p, move (nv), function<translate_function> (), iu, cd, fl, *this); + p, move (nv), function<translate_function> (), iu, cv, fl, *this); } - optional<text_type> package_manifest:: - effective_description_type (bool iu) const + strings package_manifest:: + effective_type_sub_options (const optional<string>& t) { - if (!description) - throw logic_error ("absent description"); + strings r; - optional<text_type> r; - - if (description_type) - r = to_text_type (*description_type); - else if (description->file) + if (t) { - string ext (description->path.extension ()); - if (ext.empty () || icasecmp (ext, "txt") == 0) - r = text_type::plain; - else if (icasecmp (ext, "md") == 0 || icasecmp (ext, "markdown") == 0) - r = text_type::github_mark; + for (size_t b (0), e (0); next_word (*t, b, e, ','); ) + { + if (b != 0) + r.push_back (trim (string (*t, b, e - b))); + } } - else - r = text_type::plain; - - if (!r && !iu) - throw invalid_argument ("unknown text type"); return r; } - void package_manifest:: - override (const vector<manifest_name_value>& nvs, const string& name) + // If validate_only is true, then the package manifest is assumed to be + // default constructed and is used as a storage for convenience of the + // validation implementation. + // + static void + override (const vector<manifest_name_value>& nvs, + const string& name, + package_manifest& m, + bool validate_only) { - // Reset the build constraints value sub-group on the first call. + // The first {builds, build-{include,exclude}} override value. // - bool rbc (true); - auto reset_build_constraints = [&rbc, this] () - { - if (rbc) - { - build_constraints.clear (); - rbc = false; - } - }; + const manifest_name_value* cbc (nullptr); - // Reset the builds value group on the first call. + // The first builds override value. // - bool rb (true); - auto reset_builds = [&rb, &reset_build_constraints, this] () - { - if (rb) - { - builds.clear (); - reset_build_constraints (); - rb = false; - } - }; + const manifest_name_value* cb (nullptr); + + // The first {*-builds, *-build-{include,exclude}} override value. + // + const manifest_name_value* pbc (nullptr); + + // The first {build-bot} override value. + // + const manifest_name_value* cbb (nullptr); + + // The first {*-build-bot} override value. + // + const manifest_name_value* pbb (nullptr); + + // The first {build-*email} override value. + // + const manifest_name_value* cbe (nullptr); + + // The first {*-build-*email} override value. + // + const manifest_name_value* pbe (nullptr); + + // List of indexes of the build configurations with the overridden build + // constraints together with flags which indicate if the *-builds override + // value was encountered for this configuration. + // + vector<pair<size_t, bool>> obcs; + + // List of indexes of the build configurations with the overridden bots. + // + vector<size_t> obbs; + + // List of indexes of the build configurations with the overridden emails. + // + vector<size_t> obes; - // Reset the build emails value group on the first call. + // Return true if the specified package build configuration is newly + // created by the *-build-config override. // - bool rbe (true); - auto reset_build_emails = [&rbe, this] () + auto config_created = [&m, confs_num = m.build_configs.size ()] + (const build_package_config& c) { - if (rbe) - { - build_email = nullopt; - build_warning_email = nullopt; - build_error_email = nullopt; - rbe = false; - } + return &c >= m.build_configs.data () + confs_num; }; + // Apply overrides. + // for (const manifest_name_value& nv: nvs) { + auto bad_name = [&name, &nv] (const string& d) + { + throw !name.empty () + ? parsing (name, nv.name_line, nv.name_column, d) + : parsing (d); + }; + + // Reset the build-{include,exclude} value sub-group on the first call + // but throw if any of the {*-builds, *-build-{include,exclude}} + // override values are already encountered. + // + auto reset_build_constraints = [&cbc, &pbc, &nv, &bad_name, &m] () + { + if (cbc == nullptr) + { + if (pbc != nullptr) + bad_name ('\'' + nv.name + "' override specified together with '" + + pbc->name + "' override"); + + m.build_constraints.clear (); + cbc = &nv; + } + }; + + // Reset the {builds, build-{include,exclude}} value group on the first + // call. + // + auto reset_builds = [&cb, &nv, &reset_build_constraints, &m] () + { + if (cb == nullptr) + { + reset_build_constraints (); + + m.builds.clear (); + cb = &nv; + } + }; + + // Return the reference to the package build configuration which matches + // the build config value override, if exists. If no configuration + // matches, then create one, if requested, and throw manifest_parsing + // otherwise. + // + // The n argument specifies the length of the configuration name in + // *-build-config, *-builds, *-build-{include,exclude}, *-build-bot, and + // *-build-*email values. + // + auto build_conf = + [&nv, &bad_name, &m] (size_t n, bool create) -> build_package_config& + { + const string& nm (nv.name); + small_vector<build_package_config, 1>& cs (m.build_configs); + + // Find the build package configuration. If no configuration is found, + // then create one, if requested, and throw otherwise. + // + auto i (find_if (cs.begin (), cs.end (), + [&nm, n] (const build_package_config& c) + {return nm.compare (0, n, c.name) == 0;})); + + if (i == cs.end ()) + { + string cn (nm, 0, n); + + if (create) + { + cs.emplace_back (move (cn)); + return cs.back (); + } + else + bad_name ("cannot override '" + nm + "' value: no build " + + "package configuration '" + cn + '\''); + } + + return *i; + }; + + // Return the reference to the package build configuration which matches + // the build config-specific builds group value override, if exists. If + // no configuration matches, then throw manifest_parsing, except for the + // validate-only mode in which case just add an empty configuration with + // this name and return the reference to it. Also verify that no common + // build constraints group value overrides are applied yet and throw if + // that's not the case. + // + auto build_conf_constr = + [&pbc, &cbc, &nv, &obcs, &bad_name, &build_conf, &m, validate_only] + (size_t n) -> build_package_config& + { + const string& nm (nv.name); + + // If this is the first build config override value, then save its + // address. But first verify that no common build constraints group + // value overrides are applied yet and throw if that's not the case. + // + if (pbc == nullptr) + { + if (cbc != nullptr) + bad_name ('\'' + nm + "' override specified together with '" + + cbc->name + "' override"); + + pbc = &nv; + } + + small_vector<build_package_config, 1>& cs (m.build_configs); + + // Find the build package configuration. If there is no such a + // configuration then throw, except for the validate-only mode in + // which case just add an empty configuration with this name. + // + // Note that we are using indexes rather then configuration addresses + // due to potential reallocations. + // + build_package_config& r (build_conf (n, validate_only)); + size_t ci (&r - cs.data ()); + bool bv (nm.compare (n, nm.size () - n, "-builds") == 0); + + // If this is the first encountered + // {*-builds, *-build-{include,exclude}} override for this build + // config, then clear this config' constraints member and add an entry + // to the overridden configs list. + // + auto i (find_if (obcs.begin (), obcs.end (), + [ci] (const auto& c) {return c.first == ci;})); + + bool first (i == obcs.end ()); + + if (first) + { + r.constraints.clear (); + + obcs.push_back (make_pair (ci, bv)); + } + + // If this is the first encountered *-builds override, then also clear + // this config' builds member. + // + if (bv && (first || !i->second)) + { + r.builds.clear (); + + if (!first) + i->second = true; + } + + return r; + }; + + // Reset the {build-bot} value group on the first call but throw if any + // of the {*-build-bot} override values are already encountered. + // + auto reset_build_bots = [&cbb, &pbb, &nv, &bad_name, &m] () + { + if (cbb == nullptr) + { + if (pbb != nullptr) + bad_name ('\'' + nv.name + "' override specified together with '" + + pbb->name + "' override"); + + m.build_bot_keys.clear (); + cbb = &nv; + } + }; + + // Return the reference to the package build configuration which matches + // the build config-specific build bot value override, if exists. If no + // configuration matches, then throw manifest_parsing, except for the + // validate-only mode in which case just add an empty configuration with + // this name and return the reference to it. Also verify that no common + // build bot value overrides are applied yet and throw if that's not the + // case. + // + auto build_conf_bot = + [&pbb, &cbb, &nv, &obbs, &bad_name, &build_conf, &m, validate_only] + (size_t n) -> build_package_config& + { + const string& nm (nv.name); + + // If this is the first build config override value, then save its + // address. But first verify that no common build bot value overrides + // are applied yet and throw if that's not the case. + // + if (pbb == nullptr) + { + if (cbb != nullptr) + bad_name ('\'' + nm + "' override specified together with '" + + cbb->name + "' override"); + + pbb = &nv; + } + + small_vector<build_package_config, 1>& cs (m.build_configs); + + // Find the build package configuration. If there is no such a + // configuration then throw, except for the validate-only mode in + // which case just add an empty configuration with this name. + // + // Note that we are using indexes rather then configuration addresses + // due to potential reallocations. + // + build_package_config& r (build_conf (n, validate_only)); + size_t ci (&r - cs.data ()); + + // If this is the first encountered {*-build-bot} override for this + // build config, then clear this config' bot_keys members and add an + // entry to the overridden configs list. + // + if (find (obbs.begin (), obbs.end (), ci) == obbs.end ()) + { + r.bot_keys.clear (); + + obbs.push_back (ci); + } + + return r; + }; + + // Reset the {build-*email} value group on the first call but throw if + // any of the {*-build-*email} override values are already encountered. + // + auto reset_build_emails = [&cbe, &pbe, &nv, &bad_name, &m] () + { + if (cbe == nullptr) + { + if (pbe != nullptr) + bad_name ('\'' + nv.name + "' override specified together with '" + + pbe->name + "' override"); + + m.build_email = nullopt; + m.build_warning_email = nullopt; + m.build_error_email = nullopt; + cbe = &nv; + } + }; + + // Return the reference to the package build configuration which matches + // the build config-specific emails group value override, if exists. If + // no configuration matches, then throw manifest_parsing, except for the + // validate-only mode in which case just add an empty configuration with + // this name and return the reference to it. Also verify that no common + // build emails group value overrides are applied yet and throw if + // that's not the case. + // + auto build_conf_email = + [&pbe, &cbe, &nv, &obes, &bad_name, &build_conf, &m, validate_only] + (size_t n) -> build_package_config& + { + const string& nm (nv.name); + + // If this is the first build config override value, then save its + // address. But first verify that no common build emails group value + // overrides are applied yet and throw if that's not the case. + // + if (pbe == nullptr) + { + if (cbe != nullptr) + bad_name ('\'' + nm + "' override specified together with '" + + cbe->name + "' override"); + + pbe = &nv; + } + + small_vector<build_package_config, 1>& cs (m.build_configs); + + // Find the build package configuration. If there is no such a + // configuration then throw, except for the validate-only mode in + // which case just add an empty configuration with this name. + // + // Note that we are using indexes rather then configuration addresses + // due to potential reallocations. + // + build_package_config& r (build_conf (n, validate_only)); + size_t ci (&r - cs.data ()); + + // If this is the first encountered {*-build-*email} override for this + // build config, then clear this config' email members and add an + // entry to the overridden configs list. + // + if (find (obes.begin (), obes.end (), ci) == obes.end ()) + { + r.email = nullopt; + r.warning_email = nullopt; + r.error_email = nullopt; + + obes.push_back (ci); + } + + return r; + }; + + // Parse the [*-]build-auxiliary[-*] value override. If the mode is not + // validate-only, then override the matching value and throw + // manifest_parsing if no match. But throw only unless this is a + // configuration-specific override (build_config is not NULL) for a + // newly created configuration, in which case add the value instead. + // + auto override_build_auxiliary = + [&bad_name, + &name, + &config_created, + validate_only] (const name_value& nv, + string&& en, + vector<build_auxiliary>& r, + build_package_config* build_config = nullptr) + { + build_auxiliary a (bpkg::parse_build_auxiliary (nv, move (en), name)); + + if (!validate_only) + { + auto i (find_if (r.begin (), r.end (), + [&a] (const build_auxiliary& ba) + { + return ba.environment_name == a.environment_name; + })); + + if (i != r.end ()) + { + *i = move (a); + } + else + { + if (build_config != nullptr && config_created (*build_config)) + r.emplace_back (move (a)); + else + bad_name ("no match for '" + nv.name + "' value override"); + } + } + }; + const string& n (nv.name); if (n == "builds") { reset_builds (); - builds.push_back (parse_build_class_expr (nv, builds.empty (), name)); + + m.builds.push_back ( + parse_build_class_expr (nv, m.builds.empty (), name)); } else if (n == "build-include") { reset_build_constraints (); - build_constraints.push_back ( + m.build_constraints.push_back ( parse_build_constraint (nv, false /* exclusion */, name)); } else if (n == "build-exclude") { reset_build_constraints (); - build_constraints.push_back ( + m.build_constraints.push_back ( + parse_build_constraint (nv, true /* exclusion */, name)); + } + else if (n == "build-bot") + { + reset_build_bots (); + + parse_build_bot (nv, name, m.build_bot_keys); + } + else if ((n.size () > 13 && + n.compare (n.size () - 13, 13, "-build-config") == 0)) + { + build_package_config& bc ( + build_conf (n.size () - 13, true /* create */)); + + auto vc (parser::split_comment (nv.value)); + + bc.arguments = move (vc.first); + bc.comment = move (vc.second); + } + else if (n.size () > 7 && n.compare (n.size () - 7, 7, "-builds") == 0) + { + build_package_config& bc (build_conf_constr (n.size () - 7)); + + bc.builds.push_back ( + parse_build_class_expr (nv, bc.builds.empty (), name)); + } + else if (n.size () > 14 && + n.compare (n.size () - 14, 14, "-build-include") == 0) + { + build_package_config& bc (build_conf_constr (n.size () - 14)); + + bc.constraints.push_back ( + parse_build_constraint (nv, false /* exclusion */, name)); + } + else if (n.size () > 14 && + n.compare (n.size () - 14, 14, "-build-exclude") == 0) + { + build_package_config& bc (build_conf_constr (n.size () - 14)); + + bc.constraints.push_back ( parse_build_constraint (nv, true /* exclusion */, name)); } + else if (n.size () > 10 && + n.compare (n.size () - 10, 10, "-build-bot") == 0) + { + build_package_config& bc (build_conf_bot (n.size () - 10)); + parse_build_bot (nv, name, bc.bot_keys); + } else if (n == "build-email") { reset_build_emails (); - build_email = parse_email (nv, "build", name, true /* empty */); + m.build_email = parse_email (nv, "build", name, true /* empty */); } else if (n == "build-warning-email") { reset_build_emails (); - build_warning_email = parse_email (nv, "build warning", name); + m.build_warning_email = parse_email (nv, "build warning", name); } else if (n == "build-error-email") { reset_build_emails (); - build_error_email = parse_email (nv, "build error", name); + m.build_error_email = parse_email (nv, "build error", name); + } + else if (n.size () > 12 && + n.compare (n.size () - 12, 12, "-build-email") == 0) + { + build_package_config& bc (build_conf_email (n.size () - 12)); + + bc.email = parse_email ( + nv, "build configuration", name, true /* empty */); + } + else if (n.size () > 20 && + n.compare (n.size () - 20, 20, "-build-warning-email") == 0) + { + build_package_config& bc (build_conf_email (n.size () - 20)); + + bc.warning_email = parse_email ( + nv, "build configuration warning", name); + } + else if (n.size () > 18 && + n.compare (n.size () - 18, 18, "-build-error-email") == 0) + { + build_package_config& bc (build_conf_email (n.size () - 18)); + + bc.error_email = parse_email (nv, "build configuration error", name); + } + else if (optional<pair<string, string>> ba = + build_auxiliary::parse_value_name (n)) + { + if (ba->first.empty ()) // build-auxiliary*? + { + override_build_auxiliary (nv, move (ba->second), m.build_auxiliaries); + } + else // *-build-auxiliary* + { + build_package_config& bc ( + build_conf (ba->first.size (), validate_only)); + + override_build_auxiliary (nv, move (ba->second), bc.auxiliaries, &bc); + } } else + bad_name ("cannot override '" + n + "' value"); + } + + // Common build constraints and build config overrides are mutually + // exclusive. + // + assert (cbc == nullptr || pbc == nullptr); + + // Now, if not in the validate-only mode, as all the potential build + // constraint, bot keys, and email overrides are applied, perform the + // final adjustments to the build config constraints, bot keys, and + // emails. + // + if (!validate_only) + { + if (cbc != nullptr) // Common build constraints are overridden? + { + for (build_package_config& c: m.build_configs) + { + c.builds.clear (); + c.constraints.clear (); + } + } + else if (pbc != nullptr) // Build config constraints are overridden? { - string d ("cannot override '" + n + "' value"); + for (size_t i (0); i != m.build_configs.size (); ++i) + { + if (find_if (obcs.begin (), obcs.end (), + [i] (const auto& pc) {return pc.first == i;}) == + obcs.end ()) + { + build_package_config& c (m.build_configs[i]); - throw !name.empty () - ? parsing (name, nv.name_line, nv.name_column, d) - : parsing (d); + c.builds.clear (); + c.constraints.clear (); + c.builds.emplace_back ("none", "" /* comment */); + } + } + } + + if (cbb != nullptr) // Common build bots are overridden? + { + for (build_package_config& c: m.build_configs) + c.bot_keys.clear (); + } + + if (cbe != nullptr) // Common build emails are overridden? + { + for (build_package_config& c: m.build_configs) + { + c.email = nullopt; + c.warning_email = nullopt; + c.error_email = nullopt; + } + } + else if (pbe != nullptr) // Build config emails are overridden? + { + for (size_t i (0); i != m.build_configs.size (); ++i) + { + if (find (obes.begin (), obes.end (), i) == obes.end ()) + { + build_package_config& c (m.build_configs[i]); + + c.email = email (); + c.warning_email = nullopt; + c.error_email = nullopt; + } + } } } } void package_manifest:: + override (const vector<manifest_name_value>& nvs, const string& name) + { + bpkg::override (nvs, name, *this, false /* validate_only */); + } + + void package_manifest:: validate_overrides (const vector<manifest_name_value>& nvs, const string& name) { package_manifest p; - p.override (nvs, name); + bpkg::override (nvs, name, p, true /* validate_only */); } - static const string description_file ("description-file"); - static const string changes_file ("changes-file"); + static const string description_file ("description-file"); + static const string package_description_file ("package-description-file"); + static const string changes_file ("changes-file"); + static const string build_file ("build-file"); void package_manifest:: load_files (const function<load_function>& loader, bool iu) { - auto load = [&loader] (const string& n, const path& p) + // If required, load a file and verify that its content is not empty, if + // the loader returns the content. Make the text type explicit. + // + auto load = [iu, &loader] (typed_text_file& text, + const string& file_value_name) { - string r (loader (n, p)); + // Make the type explicit. + // + optional<text_type> t; - if (r.empty ()) - throw parsing ("package " + n + " references empty file"); + // Convert the potential invalid_argument exception to the + // manifest_parsing exception similar to what we do in the manifest + // parser. + // + try + { + t = text.effective_type (iu); + } + catch (const invalid_argument& e) + { + if (text.type) + { + // Strip trailing "-file". + // + string prefix (file_value_name, 0, file_value_name.size () - 5); - return r; - }; + throw parsing ("invalid " + prefix + "-type package manifest " + + "value: " + e.what ()); + } + else + { + throw parsing ("invalid " + file_value_name + " package " + + "manifest value: " + e.what ()); + } + } - // Load the description-file manifest value. - // - if (description) - { - // Make the description type explicit. - // - optional<text_type> t (effective_description_type (iu)); // Can throw. assert (t || iu); // Can only be absent if we ignore unknown. - if (!description_type && t) - description_type = to_string (*t); + if (!text.type && t) + text.type = to_string (*t); - // At this point the description type can only be absent if the - // description comes from a file. Otherwise, we would end up with the - // plain text. + // At this point the type can only be absent if the text comes from a + // file. Otherwise, we would end up with the plain text. // - assert (description_type || description->file); + assert (text.type || text.file); - if (description->file) + if (text.file) { - if (!description_type) - description_type = "text/unknown; extension=" + - description->path.extension (); + if (!text.type) + text.type = "text/unknown; extension=" + text.path.extension (); - description = text_file (load (description_file, description->path)); + if (optional<string> fc = loader (file_value_name, text.path)) + { + if (fc->empty ()) + throw parsing ("package manifest value " + file_value_name + + " references empty file"); + + text = typed_text_file (move (*fc), move (text.type)); + } } - } + }; - // Load the changes-file manifest values. + // Load the descriptions and changes, if present. // - for (text_file& c: changes) + if (description) + load (*description, description_file); + + if (package_description) + load (*package_description, package_description_file); + + for (typed_text_file& c: changes) + load (c, changes_file); + + // Load the build-file manifest values. + // + if (!buildfile_paths.empty ()) { - if (c.file) - c = text_file (load (changes_file, c.path)); + // Must already be set if the build-file value is parsed. + // + assert (alt_naming); + + dir_path d (*alt_naming ? "build2" : "build"); + + for (auto i (buildfile_paths.begin ()); i != buildfile_paths.end (); ) + { + path& p (*i); + path f (d / p); + f += *alt_naming ? ".build2" : ".build"; + + if (optional<string> fc = loader (build_file, f)) + { + buildfiles.emplace_back (move (p), move (*fc)); + i = buildfile_paths.erase (i); // Moved to buildfiles. + } + else + ++i; + } } } static void - serialize_package_manifest (manifest_serializer& s, - const package_manifest& m, - bool header_only) + serialize_package_manifest ( + manifest_serializer& s, + const package_manifest& m, + bool header_only, + const optional<standard_version>& min_ver = nullopt) { // @@ Should we check that all non-optional values are specified ? // @@ Should we check that values are valid: version release is not empty, @@ -2645,6 +5668,12 @@ namespace bpkg if (m.upstream_version) s.next ("upstream-version", *m.upstream_version); + if (m.type) + s.next ("type", *m.type); + + for (const language& l: m.languages) + s.next ("language", !l.impl ? l.name : l.name + "=impl"); + if (m.project) s.next ("project", m.project->string ()); @@ -2672,26 +5701,46 @@ namespace bpkg if (!m.keywords.empty ()) s.next ("keywords", concatenate (m.keywords, " ")); - if (m.description) + auto serialize_text_file = [&s] (const text_file& v, const string& n) { - if (m.description->file) - s.next ("description-file", - serializer::merge_comment (m.description->path.string (), - m.description->comment)); + if (v.file) + s.next (n + "-file", + serializer::merge_comment (v.path.string (), v.comment)); else - s.next ("description", m.description->text); + s.next (n, v.text); + }; - if (m.description_type) - s.next ("description-type", *m.description_type); - } + auto serialize_description = [&s, &serialize_text_file] + (const optional<typed_text_file>& desc, + const char* prefix) + { + if (desc) + { + string p (prefix); + serialize_text_file (*desc, p + "description"); + + if (desc->type) + s.next (p + "description-type", *desc->type); + } + }; + + serialize_description (m.description, "" /* prefix */); + serialize_description (m.package_description, "package-"); for (const auto& c: m.changes) + serialize_text_file (c, "changes"); + + // If there are any changes, then serialize the type of the first + // changes entry, if present. Note that if it is present, then we assume + // that the type was specified explicitly and so it is the same for all + // entries. + // + if (!m.changes.empty ()) { - if (c.file) - s.next ("changes-file", - serializer::merge_comment (c.path.string (), c.comment)); - else - s.next ("changes", c.text); + const typed_text_file& c (m.changes.front ()); + + if (c.type) + s.next ("changes-type", *c.type); } if (m.url) @@ -2738,21 +5787,29 @@ namespace bpkg m.build_error_email->comment)); for (const dependency_alternatives& d: m.dependencies) - s.next ("depends", - (d.conditional - ? (d.buildtime ? "?* " : "? ") - : (d.buildtime ? "* " : "")) + - serializer::merge_comment (concatenate (d, " | "), d.comment)); + s.next ("depends", d.string ()); for (const requirement_alternatives& r: m.requirements) - s.next ("requires", - (r.conditional - ? (r.buildtime ? "?* " : "? ") - : (r.buildtime ? "* " : "")) + - serializer::merge_comment (concatenate (r, " | "), r.comment)); + s.next ("requires", r.string ()); + + for (const test_dependency& t: m.tests) + { + string n (to_string (t.type)); - for (const test_dependency& p: m.tests) - s.next (to_string (p.type), p.string ()); + // If we generate the manifest for parsing by clients of libbpkg + // versions less than 0.14.0-, then replace the introduced in 0.14.0 + // build-time tests, examples, and benchmarks values with + // tests-0.14.0, examples-0.14.0, and benchmarks-0.14.0, + // respectively. This way such clients will still be able to parse it, + // ignoring unknown values. + // + // @@ TMP time to drop? + // 0.14.0- + if (t.buildtime && min_ver && min_ver->version < 13999990001ULL) + n += "-0.14.0"; + + s.next (n, t.string ()); + } for (const build_class_expr& e: m.builds) s.next ("builds", serializer::merge_comment (e.string (), e.comment)); @@ -2761,9 +5818,97 @@ namespace bpkg s.next (c.exclusion ? "build-exclude" : "build-include", serializer::merge_comment (!c.target ? c.config - : c.config + "/" + *c.target, + : c.config + '/' + *c.target, c.comment)); + for (const build_auxiliary& ba: m.build_auxiliaries) + s.next ((!ba.environment_name.empty () + ? "build-auxiliary-" + ba.environment_name + : "build-auxiliary"), + serializer::merge_comment (ba.config, ba.comment)); + + for (const string& k: m.build_bot_keys) + s.next ("build-bot", k); + + for (const build_package_config& bc: m.build_configs) + { + if (!bc.builds.empty ()) + { + string n (bc.name + "-builds"); + for (const build_class_expr& e: bc.builds) + s.next (n, serializer::merge_comment (e.string (), e.comment)); + } + + if (!bc.constraints.empty ()) + { + string in (bc.name + "-build-include"); + string en (bc.name + "-build-exclude"); + + for (const build_constraint& c: bc.constraints) + s.next (c.exclusion ? en : in, + serializer::merge_comment (!c.target + ? c.config + : c.config + '/' + *c.target, + c.comment)); + } + + if (!bc.auxiliaries.empty ()) + { + string n (bc.name + "-build-auxiliary"); + + for (const build_auxiliary& ba: bc.auxiliaries) + s.next ((!ba.environment_name.empty () + ? n + '-' + ba.environment_name + : n), + serializer::merge_comment (ba.config, ba.comment)); + } + + if (!bc.bot_keys.empty ()) + { + string n (bc.name + "-build-bot"); + + for (const string& k: bc.bot_keys) + s.next (n, k); + } + + if (!bc.arguments.empty () || !bc.comment.empty ()) + s.next (bc.name + "-build-config", + serializer::merge_comment (bc.arguments, bc.comment)); + + if (bc.email) + s.next (bc.name + "-build-email", + serializer::merge_comment (*bc.email, bc.email->comment)); + + if (bc.warning_email) + s.next (bc.name + "-build-warning-email", + serializer::merge_comment (*bc.warning_email, + bc.warning_email->comment)); + + if (bc.error_email) + s.next (bc.name + "-build-error-email", + serializer::merge_comment (*bc.error_email, + bc.error_email->comment)); + } + + bool an (m.alt_naming && *m.alt_naming); + + if (m.bootstrap_build) + s.next (an ? "bootstrap-build2" : "bootstrap-build", + *m.bootstrap_build); + + if (m.root_build) + s.next (an ? "root-build2" : "root-build", *m.root_build); + + for (const auto& bf: m.buildfiles) + s.next (bf.path.posix_string () + (an ? "-build2" : "-build"), + bf.content); + + for (const path& f: m.buildfile_paths) + s.next ("build-file", f.posix_string () + (an ? ".build2" : ".build")); + + for (const distribution_name_value& nv: m.distribution_values) + s.next (nv.name, nv.value); + if (m.location) s.next ("location", m.location->posix_string ()); @@ -2778,9 +5923,9 @@ namespace bpkg } void package_manifest:: - serialize (serializer& s) const + serialize (serializer& s, const optional<standard_version>& min_ver) const { - serialize_package_manifest (s, *this, false); + serialize_package_manifest (s, *this, false, min_ver); } void package_manifest:: @@ -2994,7 +6139,7 @@ namespace bpkg } void pkg_package_manifests:: - serialize (serializer& s) const + serialize (serializer& s, const optional<standard_version>& min_ver) const { // Serialize the package list manifest. // @@ -3012,21 +6157,32 @@ namespace bpkg { throw serialization ( s.name (), - d + " for " + p.name.string () + "-" + p.version.string ()); + d + " for " + p.name.string () + '-' + p.version.string ()); }; - if (p.description) + // Throw manifest_serialization if the text is in a file or untyped. + // + auto verify_text_file = [&bad_value] (const typed_text_file& v, + const string& n) { - if (p.description->file) - bad_value ("forbidden description-file"); + if (v.file) + bad_value ("forbidden " + n + "-file"); - if (!p.description_type) - bad_value ("no valid description-type"); - } + if (!v.type) + bad_value ("no valid " + n + "-type"); + }; + + if (p.description) + verify_text_file (*p.description, "description"); + + if (p.package_description) + verify_text_file (*p.package_description, "package-description"); for (const auto& c: p.changes) - if (c.file) - bad_value ("forbidden changes-file"); + verify_text_file (c, "changes"); + + if (!p.buildfile_paths.empty ()) + bad_value ("forbidden build-file"); if (!p.location) bad_value ("no valid location"); @@ -3034,7 +6190,7 @@ namespace bpkg if (!p.sha256sum) bad_value ("no valid sha256sum"); - pkg_package_manifest (s, p); + pkg_package_manifest (s, p, min_ver); } s.next ("", ""); // End of stream. @@ -3380,7 +6536,7 @@ namespace bpkg if (optional<repository_type> r = parse_repository_type (t)) return *r; - throw invalid_argument ("invalid repository type '" + t + "'"); + throw invalid_argument ("invalid repository type '" + t + '\''); } repository_type @@ -4039,7 +7195,8 @@ namespace bpkg parse_repository_manifest (parser& p, name_value nv, repository_type base_type, - bool iu) + bool iu, + bool verify_version = true) { auto bad_name ([&p, &nv](const string& d) { throw parsing (p.name (), nv.name_line, nv.name_column, d);}); @@ -4049,11 +7206,16 @@ namespace bpkg // Make sure this is the start and we support the version. // - if (!nv.name.empty ()) - bad_name ("start of repository manifest expected"); + if (verify_version) + { + if (!nv.name.empty ()) + bad_name ("start of repository manifest expected"); - if (nv.value != "1") - bad_value ("unsupported format version"); + if (nv.value != "1") + bad_value ("unsupported format version"); + + nv = p.next (); + } repository_manifest r; @@ -4064,7 +7226,7 @@ namespace bpkg optional<repository_type> type; optional<name_value> location; - for (nv = p.next (); !nv.empty (); nv = p.next ()) + for (; !nv.empty (); nv = p.next ()) { string& n (nv.name); string& v (nv.value); @@ -4409,13 +7571,126 @@ namespace bpkg parse_repository_manifests (parser& p, repository_type base_type, bool iu, + optional<repositories_manifest_header>& header, vector<repository_manifest>& ms) { + // Return nullopt on eos. Otherwise, parse and verify the + // manifest-starting format version value and return the subsequent + // manifest value, that can potentially be empty (for an empty manifest). + // + // Also save the manifest-starting position (start_nv) for the + // diagnostics. + // + name_value start_nv; + auto next_manifest = [&p, &start_nv] () -> optional<name_value> + { + start_nv = p.next (); + + if (start_nv.empty ()) + return nullopt; + + // Make sure this is the start and we support the version. + // + if (!start_nv.name.empty ()) + throw parsing (p.name (), start_nv.name_line, start_nv.name_column, + "start of repository manifest expected"); + + if (start_nv.value != "1") + throw parsing (p.name (), start_nv.value_line, start_nv.value_column, + "unsupported format version"); + + return p.next (); + }; + + optional<name_value> nv (next_manifest ()); + + if (!nv) + throw parsing (p.name (), start_nv.name_line, start_nv.name_column, + "start of repository manifest expected"); + + auto bad_name ([&p, &nv](const string& d) { + throw parsing (p.name (), nv->name_line, nv->name_column, d);}); + + auto bad_value ([&p, &nv](const string& d) { + throw parsing (p.name (), nv->value_line, nv->value_column, d);}); + + // First check if this a header manifest, if any manifest is present. + // + // Note that if this is none of the known header values, then we assume + // this is a repository manifest (rather than a header that starts with an + // unknown value; so use one of the original names to make sure it's + // recognized as such, for example `compression:none`). + // + if (nv->name == "min-bpkg-version" || + nv->name == "compression") + { + header = repositories_manifest_header (); + + // First verify the version, if any. + // + if (nv->name == "min-bpkg-version") + try + { + const string& v (nv->value); + standard_version mbv (v, standard_version::allow_earliest); + + if (mbv > standard_version (LIBBPKG_VERSION_STR)) + bad_value ( + "incompatible repositories manifest: minimum bpkg version is " + v); + + header->min_bpkg_version = move (mbv); + + nv = p.next (); + } + catch (const invalid_argument& e) + { + bad_value (string ("invalid minimum bpkg version: ") + e.what ()); + } + + // Parse the remaining header values, failing if min-bpkg-version is + // encountered (should be first). + // + for (; !nv->empty (); nv = p.next ()) + { + const string& n (nv->name); + string& v (nv->value); + + if (n == "min-bpkg-version") + { + bad_name ("minimum bpkg version must be first in repositories " + "manifest header"); + } + else if (n == "compression") + { + header->compression = move (v); + } + else if (!iu) + bad_name ("unknown name '" + n + "' in repositories manifest header"); + } + + nv = next_manifest (); + } + + // Parse the manifest list. + // + // Note that if nv is present, then it contains the manifest's first + // value, which can potentially be empty (for an empty manifest, which is + // recognized as a base manifest). + // + // Also note that if the header is present but is not followed by + // repository manifests (there is no ':' line after the header values), + // then the empty manifest list is returned (no base manifest is + // automatically added). + // bool base (false); - for (name_value nv (p.next ()); !nv.empty (); nv = p.next ()) + while (nv) { - ms.push_back (parse_repository_manifest (p, nv, base_type, iu)); + ms.push_back (parse_repository_manifest (p, + *nv, + base_type, + iu, + false /* verify_version */)); // Make sure that there is a single base repository manifest in the // list. @@ -4423,19 +7698,38 @@ namespace bpkg if (ms.back ().effective_role () == repository_role::base) { if (base) - throw parsing (p.name (), nv.name_line, nv.name_column, + throw parsing (p.name (), start_nv.name_line, start_nv.name_column, "base repository manifest redefinition"); base = true; } + + nv = next_manifest (); } } // Serialize the repository manifest list. // static void - serialize_repository_manifests (serializer& s, - const vector<repository_manifest>& ms) + serialize_repository_manifests ( + serializer& s, + const optional<repositories_manifest_header>& header, + const vector<repository_manifest>& ms) { + if (header) + { + s.next ("", "1"); // Start of manifest. + + const repositories_manifest_header& h (*header); + + if (h.min_bpkg_version) + s.next ("min-bpkg-version", h.min_bpkg_version->string ()); + + if (h.compression) + s.next ("compression", *h.compression); + + s.next ("", ""); // End of manifest. + } + for (const repository_manifest& r: ms) r.serialize (s); @@ -4447,13 +7741,13 @@ namespace bpkg pkg_repository_manifests:: pkg_repository_manifests (parser& p, bool iu) { - parse_repository_manifests (p, repository_type::pkg, iu, *this); + parse_repository_manifests (p, repository_type::pkg, iu, header, *this); } void pkg_repository_manifests:: serialize (serializer& s) const { - serialize_repository_manifests (s, *this); + serialize_repository_manifests (s, header, *this); } // dir_repository_manifests @@ -4461,13 +7755,13 @@ namespace bpkg dir_repository_manifests:: dir_repository_manifests (parser& p, bool iu) { - parse_repository_manifests (p, repository_type::dir, iu, *this); + parse_repository_manifests (p, repository_type::dir, iu, header, *this); } void dir_repository_manifests:: serialize (serializer& s) const { - serialize_repository_manifests (s, *this); + serialize_repository_manifests (s, header, *this); } // git_repository_manifests @@ -4475,13 +7769,13 @@ namespace bpkg git_repository_manifests:: git_repository_manifests (parser& p, bool iu) { - parse_repository_manifests (p, repository_type::git, iu, *this); + parse_repository_manifests (p, repository_type::git, iu, header, *this); } void git_repository_manifests:: serialize (serializer& s) const { - serialize_repository_manifests (s, *this); + serialize_repository_manifests (s, header, *this); } // signature_manifest @@ -4601,13 +7895,13 @@ namespace bpkg } version - extract_package_version (const char* s, bool fold_zero_revision) + extract_package_version (const char* s, version::flags fl) { using traits = string::traits_type; if (const char* p = traits::find (s, traits::length (s), '/')) { - version r (p + 1, fold_zero_revision); + version r (p + 1, fl); if (r.release && r.release->empty ()) throw invalid_argument ("earliest version"); |