summaryrefslogtreecommitdiffstats
Commit message (Collapse)AuthorAgeFilesLines
* Bug #240684 - Fix _ebuild_arg_to_phase() to handle the src_prepare phaseZac Medico2008-10-091-0/+3
| | | | | | so that the default() function is properly created. svn path=/main/trunk/; revision=11664
* In EbuildFetcher._set_returncode(), don't clean up PORTAGE_BUILDDIR ifZac Medico2008-10-091-1/+3
| | | | | | keepwork or keeptemp are in FEATURES. svn path=/main/trunk/; revision=11661
* Don't show the repo display when the merge list is empty.Zac Medico2008-10-081-1/+1
| | | | svn path=/main/trunk/; revision=11660
* Fix apparent breakage from r11593 (slot dep support):Zac Medico2008-10-081-6/+22
| | | | | | | | * Handle KeyError from element.getAttribute() in makeAtom() and makeVersion(). * Avoid 'sre_constants.error: unmatched group' exceptions in revisionMatch() when the atom does not have a slot. svn path=/main/trunk/; revision=11659
* In EbuildBuild._fetch_exit(), clean up the fetch log if fetch was successful.Zac Medico2008-10-081-0/+7
| | | | svn path=/main/trunk/; revision=11658
* Use separate temp directories in order to avoid potential name collisions.Zac Medico2008-10-081-7/+8
| | | | svn path=/main/trunk/; revision=11657
* Since EbuildBuild no longer directs it's EbuildFetcher instance's outputZac Medico2008-10-081-9/+2
| | | | | | | to the global fetch log, there's no need to schedule it on the fetch queue. svn path=/main/trunk/; revision=11656
* Before returning config instances from Scheduler._allocate_config(), call theZac Medico2008-10-081-0/+4
| | | | | | | | | | config.reset() method to make sure all settings from the previous package get flushed out (such as PORTAGE_LOG_FILE). This fixes a case where PORTAGE_LOG_FILE leaked from the previous package to the clean phase of the next one, resulting in an IOError due to the path being invalid and the parent directory being nonexistent. svn path=/main/trunk/; revision=11655
* Generate elog messages for failed fetches. This restores some functionalityZac Medico2008-10-071-22/+26
| | | | | | that was removed in r11645. svn path=/main/trunk/; revision=11654
* Instead of using inheritance, reference the ConfigParser part of SetConfig asZac Medico2008-10-071-24/+30
| | | | | | | an attribute in case we want to swap out the ConfigParser implemention and no longer want to implement the whole interface. svn path=/main/trunk/; revision=11653
* Make "qawarnings" a normal set since the missingvars code might modify it.Zac Medico2008-10-071-1/+1
| | | | svn path=/main/trunk/; revision=11652
* Change the "qawarnings" variable from a list to a set.Zac Medico2008-10-071-3/+3
| | | | svn path=/main/trunk/; revision=11651
* Make HOMEPAGE.missing fatal again since, in cases when homepages are no longerZac Medico2008-10-071-1/+0
| | | | | | | available, we can use a link to some site like packages.gentoo.org or www.unmaintained-free-software.org. svn path=/main/trunk/; revision=11650
* When EbuildBuild encounters a failed fetch, propagee the log path so thatZac Medico2008-10-071-0/+2
| | | | | | it can be properly displayed in the status display. svn path=/main/trunk/; revision=11649
* Make EbuildFetcher clean up the build dir after failures in fetchonly modeZac Medico2008-10-071-1/+1
| | | | | | since the log is in emerge-fetch.log anyway. svn path=/main/trunk/; revision=11648
* Make EbuildFetcher direct fetch output to the normal build log. By keeping theZac Medico2008-10-071-10/+19
| | | | | | | | output in the build log instead of the global emerge-fetch.log file, output from a possible fetch failure can be recovered and displayed later. TODO: Convert failed fetch to elog message as suggested in bug #234301. svn path=/main/trunk/; revision=11647
* Fix EbuildFetcher._wait_hook() so it can be called more than once withoutZac Medico2008-10-071-10/+11
| | | | | | throwing an AttributeError. svn path=/main/trunk/; revision=11646
* Bug #234301 - Add support to EbuildFetcher to collect elog messages thatZac Medico2008-10-072-15/+45
| | | | | | | | might be generated during the pkg_nofetch phase. This involves maintaining a lock on PORTAGE_BUILDDIR where the elog messages are collected, and cleaning up PORTAGE_BUILDDIR before releasing the lock. svn path=/main/trunk/; revision=11645
* In fetch(), never call the pkg_nofetch phase when in "listonly" mode.Zac Medico2008-10-061-3/+1
| | | | svn path=/main/trunk/; revision=11644
* In doebuild(), don't create build dirs when "listonly" fetch mode.Zac Medico2008-10-061-1/+2
| | | | svn path=/main/trunk/; revision=11643
* In order to avoid "waiting for lock" messagesZac Medico2008-10-061-1/+4
| | | | | | | at the beginning, which annoy users, never spawn a prefetcher for the first package. svn path=/main/trunk/; revision=11642
* Fix KeyError raised from display_preserved_libs() when the realpath is usedZac Medico2008-10-061-2/+4
| | | | | | | as a key in consumer_map instead of the orignal path that was returned from the PreservedLibsRegistry. svn path=/main/trunk/; revision=11637
* Bug #240022 - Avoid duplicate output for the same library (due to symlinks)Zac Medico2008-10-051-1/+11
| | | | | | | | in display_preserved_libs() by using os.path.realpath() to group duplicate references together. Thanks to Fabian Groffen <grobian@g.o> for the initial patch. svn path=/main/trunk/; revision=11636
* Make HOMEPAGE.missing a warning since some packages become homeless evenZac Medico2008-10-051-0/+1
| | | | | | | though they still continue to work. Thanks to Robin H. Johnson <robbat2@g.o> for the suggestion. svn path=/main/trunk/; revision=11634
* Avoid raising a KeyError from display_preserved_libs() in cases when noZac Medico2008-10-051-3/+4
| | | | | | owners are found for a particular library consumer. svn path=/main/trunk/; revision=11633
* In dyn_clean, cd to $PORTAGE_BUILDDIR/.. before attempting to remove it sinceZac Medico2008-10-051-3/+4
| | | | | | | | some kernels, such as Solaris, return EINVAL when an attempt is made to remove the current working directory. Thanks to Fabian Groffen <grobian@g.o> for reporting. svn path=/main/trunk/; revision=11630
* Bug #239529 - When doins is called on a symlink to a directory, preserve theZac Medico2008-10-051-5/+17
| | | | | | | name of the symlink for the installed directory. This involves temporarily renaming the directory and then renaming it back again. svn path=/main/trunk/; revision=11629
* * When ensuring sane $PWD in ebuild.sh, die if the `cd "$PORTAGE_BUILDDIR"`Zac Medico2008-10-052-3/+5
| | | | | | | | call fails. * Create $PORTAGE_BUILDDIR for the "fetch" phase too since it might be necessary to call pkg_nofetch. svn path=/main/trunk/; revision=11628
* Only update the confmem file when something has changed.Zac Medico2008-10-051-3/+5
| | | | svn path=/main/trunk/; revision=11627
* Fix PreservedLibsRegistry.store() so that it only rewrites the data file whenZac Medico2008-10-051-2/+3
| | | | | | something has changed. svn path=/main/trunk/; revision=11626
* Redirect pushd and popd output to /dev/null when appropriate.Zac Medico2008-10-051-3/+3
| | | | svn path=/main/trunk/; revision=11625
* Add missing quotes on pushd argument from previous commit.Zac Medico2008-10-051-1/+1
| | | | svn path=/main/trunk/; revision=11624
* Bug #239529 - Fix 'doins' to work with symlinks to directories, like it didZac Medico2008-10-051-3/+11
| | | | | | before the changes from bug #210575. svn path=/main/trunk/; revision=11623
* A cpv can be returned from dbapi.match() as an old-style virtual match evenZac Medico2008-10-051-6/+25
| | | | | | | | | | | in cases when the package does not actually PROVIDE the virtual. Filter out any such false matches inside depgraph._select_package() and _show_unsatisfied_dep(). Thanks to Ned Ludd <solar@g.o> for reporting this issue which was discovered when attempting to install virtual/libc on uclibc profile. Apparently the uclibc ebuilds need to be fixed to properly set PROVIDE when the metadata cache is generated. svn path=/main/trunk/; revision=11622
* Handle AttributeError in case os.getloadavg does not exist, like when pythonZac Medico2008-10-041-2/+2
| | | | | | is built agains uclibc. svn path=/main/trunk/; revision=11621
* Fix depgraph._add_pkg() to ensure that the existing package in the graphZac Medico2008-10-041-1/+10
| | | | | | | | | | matches the required atom in cases when a different package has been passed in. This solves a problem with silently unsatisfied USE deps that was reported by Jorge Manuel B. S. Vicetto <jmbsvicetto@g.o>. Now the unsatisfied dependency will result in a "slot conflict" rather than be silently ignored. svn path=/main/trunk/; revision=11618
* Make depgraph._add_pkg() take a Dependency instance. This provides access toZac Medico2008-10-041-14/+21
| | | | | | | | the atom which pulled in the package, which may be needed in order to verify that a different package that's been added to the graph is capable of satisfying the atom. svn path=/main/trunk/; revision=11617
* Fix `svn status` output parsing for compatibility with subversion-1.6. ThanksZac Medico2008-10-021-5/+5
| | | | | | to Arfrever for this patch. svn path=/main/trunk/; revision=11613
* Remove pointless copying of the owners dict inside display_preserved_libs().Zac Medico2008-10-021-3/+1
| | | | svn path=/main/trunk/; revision=11611
* Implement a new "changelog.ebuildadded" check which causes repoman to bailZac Medico2008-10-022-16/+44
| | | | | | | | out if an ebuild has been added and the ChangeLog has not been modified. This was requested by Robin H Johnson <robbat2@g.o> since it is a requirement for the packages.gentoo.org ChangeLog code. svn path=/main/trunk/; revision=11610
* In JobStatusDisplay._update(), set _displayed = True when ruturning early dueZac Medico2008-10-021-0/+1
| | | | | | | to stdout not being a tty. This fixes a problem excess output which was reported by Daniel Robbins. svn path=/main/trunk/; revision=11609
* Optimize display_preserved_libs() by using a single getFileOwnerMap() to lookZac Medico2008-10-021-3/+17
| | | | | | up all the files. svn path=/main/trunk/; revision=11608
* Fix broken logic in Scheduler._main_loop() which would allow the main loop toZac Medico2008-10-021-1/+1
| | | | | | | exit while one or more merges are still scheduled. Thanks to Patrick Lauer for reporting. svn path=/main/trunk/; revision=11607
* Call elog_process before calling the "clean" phase for --buildpkgonly mode.Zac Medico2008-10-021-0/+1
| | | | svn path=/main/trunk/; revision=11606
* When parsing PORTAGE_ELOG_SYSTEM, stack values in a dict so that things likeZac Medico2008-10-021-5/+15
| | | | | | | "echo:info echo:qa" stack like one might expect. Thanks to Betelgeuse for reporting. svn path=/main/trunk/; revision=11605
* Bug #239272 - Handle InvalidAtom exception insideZac Medico2008-10-011-4/+10
| | | | | | depgraph.validate_blockers(). svn path=/main/trunk/; revision=11604
* Generate an eqawarn message if econf is called from src_compile andZac Medico2008-10-011-0/+4
| | | | | | | | src_configure is also defined. Thanks to Betelgeuse for the patch which I've modified to be silent in EAPI 0 or 1 in case ebuilds want to define src_configure in those EAPIs and call it manually from src_compile. svn path=/main/trunk/; revision=11603
* Handle InvalidDependString from portdbapi.getFetchMap() insideZac Medico2008-09-301-5/+11
| | | | | | search.output(). Thanks to agaffney for reporting. svn path=/main/trunk/; revision=11602
* Now that elog_base() uses 'echo -e' to expand escape codes prior to usingZac Medico2008-09-302-6/+2
| | | | | | | | 'read' to split on newlines, it's safe to use newlines as delimiters in the log file since 'read' is guaranteed to split any newlines contained in the arguments. svn path=/main/trunk/; revision=11601
* Bug #239006 - In FakeVartree._aux_get_wrapper(), fall back to vdb metadataZac Medico2008-09-291-1/+3
| | | | | | if the live ebuild's EAPI is unsupported. svn path=/main/trunk/; revision=11600
generated by cgit v1.2.3-1-g7c22 (git 2.25.1) at 2024-07-04 22:06:20 +0000