aboutsummaryrefslogtreecommitdiff
path: root/bpkg/pkg-build.cxx
diff options
context:
space:
mode:
authorBoris Kolpackov <boris@codesynthesis.com>2016-01-26 20:36:15 +0200
committerBoris Kolpackov <boris@codesynthesis.com>2016-01-26 20:36:15 +0200
commitc010ca31dfd32cc6255cfe8242302706a00a42a0 (patch)
tree9103aa2abbdb75032507c953aed04f2d010e0e2d /bpkg/pkg-build.cxx
parent733ed5b60710c64a1851ad92706c52fcf58c19f7 (diff)
Fix bugs in revision comparison
Diffstat (limited to 'bpkg/pkg-build.cxx')
-rw-r--r--bpkg/pkg-build.cxx34
1 files changed, 19 insertions, 15 deletions
diff --git a/bpkg/pkg-build.cxx b/bpkg/pkg-build.cxx
index 0d663bd..9c3914c 100644
--- a/bpkg/pkg-build.cxx
+++ b/bpkg/pkg-build.cxx
@@ -66,20 +66,20 @@ namespace bpkg
//
if (c)
{
- // If the constraint is equality and the revision is not explicitly
- // specified, then compare ignoring the revision. The idea is that when
- // the user runs 'bpkg build libfoo/1.0.0' and there is 1.0.0+1
- // available, it should just work. The user shouldn't have to spell the
- // revision explicitly. Similarly, when we have 'depends: libfoo ==
- // 1.0.0', then it would be strange if 1.0.0+1 did not satisfy this
- // constraint.
+ // If the revision is not explicitly specified, then compare ignoring the
+ // revision. The idea is that when the user runs 'bpkg build libfoo/1'
+ // and there is 1+1 available, it should just work. The user shouldn't
+ // have to spell the revision explicitly. Similarly, when we have
+ // 'depends: libfoo == 1', then it would be strange if 1+1 did not
+ // satisfy this constraint. The same for libfoo <= 1 -- 1+1 should
+ // satisfy.
//
// Note that strictly speaking 0 doesn't mean unspecified. Which means
// with this implementation there is no way to say "I really mean
- // revision 0" since 1.0.0 == 1.0.0+0. One can, in the current model, say
- // libfoo == 1.0.0+1, though. This is probably ok since one would assume
- // any subsequent revision of a package is only better than the ones
- // before.
+ // revision 0" since 1 == 1+0. One can, in the current model, say libfoo
+ // == 1+1, though. This is probably ok since one would assume any
+ // subsequent revision of a package version are just as (un)satisfactory
+ // as the first one.
//
if (c->min_version &&
c->max_version &&
@@ -96,18 +96,22 @@ namespace bpkg
{
if (c->min_version)
{
+ const version& v (*c->min_version);
+
if (c->min_open)
- q = q && vm > *c->min_version;
+ q = q && compare_version_gt (vm, v, v.revision != 0);
else
- q = q && vm >= *c->min_version;
+ q = q && compare_version_ge (vm, v, v.revision != 0);
}
if (c->max_version)
{
+ const version& v (*c->max_version);
+
if (c->max_open)
- q = q && vm < *c->max_version;
+ q = q && compare_version_lt (vm, v, v.revision != 0);
else
- q = q && vm <= *c->max_version;
+ q = q && compare_version_le (vm, v, v.revision != 0);
}
q += order_by_version_desc (vm);