| Age | Commit message (Collapse) | Author |
|---|
|
|
|
Props danielsh (Daniel Shahaf) in #zsh.
|
|
file does not normally contain configuration and it changes frequently. Closes: #651168
|
|
|
|
exists.
The file could exist because AVOID_COMMIT_BEFORE_INSTALL is set,
or just because the install was interrupted/system crashed etc.
If the file's more than a day old, continue with autocommit.
|
|
|
|
|
|
|
|
Avoid running find on /etc three times, only run once; and avoid
redundant user name lookups, and calls to stat(1). All done in perl now.
|
|
|
|
|
|
separate call to git add -u. Thanks to Miklos Vajna, whose patch in 2008 was deferred because -A was then too new, and languished in a branch until found today.
|
|
Avoid runing stat(1) thousands of times.
Note that etckeeper init is still slowed down by all this metadata, but
that should be rarely used.
|
|
Conflicts:
debian/changelog
|
|
-- Joey Hess <joeyh@debian.org> Fri, 25 Nov 2011 12:00:55 -0400
|
|
|
|
permissions of 644 and 755
This is unfortunately necessary in order to support etckeeper init on a
checkout that was made with a nonstandard umask, in which case the files
that were expected to be 644 and 755, won't be. Closes: #649701
Thanks to Дмитрий Матросов for reporting the bug and developing a fixup
script (attached to the bug) which could be used if you've already
encountered this problem.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
avoid daily noise. (gulikoza)
|
|
|
|
|
|
|
|
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=709487 Thanks, Thomas Moschny
|
|
|
|
after the files staged for commit are determined and so .etckeeper cannot be staged as part of the current commit. Closes: #621827
|
|
See #619409
|
|
|
|
Thanks Sjoerd Mullender for patch. Closes: https://bugzilla.redhat.com/show_bug.cgi?id=688991
|
|
|
|
|
|
GIT_AUTHOR_NAME, GIT_AUTHOR_EMAIL, GIT_COMMITTER_EMAIL) if they don't already exist.
|
|
|
|
|
|
git ls-files emits paths with no leading subdirectory. But the regexp
still needs to avoid false positives.
|
|
|
|
|
|
This avoids using git ls-files, and thus avoids encoding problems with
filenames.
Using git add . -u stages all deleted files for deletion.
It also avoids the problem with git rm deleting empty directories.
It might be better to just use git commit -a, but I've taken the
conservative approach.
There are a few other git ls-files uses in etckeeper that I have not fixed,
but they are only shown to the user, and shouldn't cause a crash as this did.
Unfortunatly, git does expose the nasty C-style encoding of such characters
in filenames to the user all over (eg, git commit, git log ..), so fixing
etckeeper to not also expose them is probably a waste of time. And
core.quotepath can be unset by users who want to clean that up.
Bug report: https://bugzilla.altlinux.org/show_bug.cgi?id=24903
|
|
|
|
|
|
settings in eg, crontab. Closes: #602438
|
|
|
|
ITIL compliance requires that _every_ package
change is detectable.
This fix ensures etckeeper properly reports any
change in rpm manifest, including epoch and release.
It uses an RPM query-format string that yields
reproducible results.
By example, assume you update a package that
differs only in the release number.
Before fix, etckeeper logs:
committing changes in /etc after yum run
Package changes:
After fix, etckeeper logs:
committing changes in /etc after yum run
Package changes:
-0:baseline-0.3.6-2.fc13.noarch
-0:baseline-devel-0.3.6-2.fc13.noarch
-0:baseline-release-0.3.6-2.fc13.noarch
+0:baseline-0.3.6-3.fc13.noarch
+0:baseline-devel-0.3.6-3.fc13.noarch
+0:baseline-release-0.3.6-3.fc13.noarch
Signed-off-by: Paul Morgan <jumanjiman@gmail.com>
|
|
|
|
|
