tigefa4u projects Trac

MultiRepos: Merged with 0.12dev [8633:8651].

MultiRepos: Changed `source` and `changeset` resources from having a composite `(reponame, id)` id to being children of a new `repository` resource.

Closes #7706.

0.12dev: Follow-up to [8612], where a check for `fs.list == None` has been inadvertently removed.

In Windows' transactioned rename, one shouldn't use arbitrarily large transaction description, as this can make the call fail.

Closes #8725.

Follow-up to r8648, as on Windows Vista x64, Python 2.3.5 would fail with `ValueError: empty range for randrange()`.

This is because 0xffffffff is -1, in Python 2.3:
{{{
>>> 0xffffffff
<stdin>:1: FutureWarning: hex/oct constants > sys.maxint will return positive values in Python 2.4 and up
-1
}}}

0.11-stable: Fixed `rename()` for Python 2.3. Also, added basic unit tests for `AtomicFile`.

Related to #8725.

TracMercurial for 0.12: ported r8646, 1.4pre compatibility fix from mercurial-plugin-0.11 branch

TracMercurial for 0.11: support Mercurial 1.4dev (after hg:1ef630452e0b).

Thanks to Benoit Boissinot on #mercurial for the hint about not supporting unreleased Mercurial versions.
Actually an unreleased version will always be considered to be the newest. If this assumption is wrong, upgrading to mpm's hg tip should do, or at least getting a release version so that the coarse grained version check will work.

Oops, follow-up to r8644, last minute refactorings are evil...

0.11.6dev: improve startup time by caching the component activation "rules" and using a dict to speed up the checks.

Before, the n components were checked against the n' rules, where n and n' were often of the same magnitude, resulting in O(n^2^) behavior.

Profiling data for startup sequence hinted this could be a significant slowdown.