summaryrefslogtreecommitdiffstats
path: root/pym/_emerge/__init__.py
Commit message (Collapse)AuthorAgeFilesLines
* Bug #247776 - Show a warning message if CONFIG_PROTECT is empty. (trunk r12022)Zac Medico2008-11-211-0/+9
| | | | svn path=/main/branches/2.1.6/; revision=12023
* Bug #248059 - Make --depclean more tolerant of invalid atoms in dependenciesZac Medico2008-11-211-5/+4
| | | | | | of packages that will be uninstalled anyway. (trunk r12020) svn path=/main/branches/2.1.6/; revision=12021
* Bug #248059 - In action_depclean(), ignore invalid atoms in deps of packagesZac Medico2008-11-211-1/+4
| | | | | | to be uninstalled. (trunk r12018) svn path=/main/branches/2.1.6/; revision=12019
* In EbuildProcess._start(), don't open the log file during the clean phaseZac Medico2008-11-211-1/+5
| | | | | | | | since the open file can result in an nfs lock on $T/build.log which prevents the clean phase from removing $T. Thanks to Jeremy Olexa <darkside@g.o> for reporting. (trunk r12016) svn path=/main/branches/2.1.6/; revision=12017
* In JobStatusDisplay, always flush the output stream after writing to it.Zac Medico2008-11-211-0/+4
| | | | | | (trunk r12014) svn path=/main/branches/2.1.6/; revision=12015
* When given an ambiguous ebuild name to install, format the list of choicesZac Medico2008-11-211-8/+37
| | | | | | | | in emerge --search format if --quiet mode is not enabled, otherwise just show a brief list. Thanks to Markus Meier <maekke@g.o> for the suggestion. (trunk r12012) svn path=/main/branches/2.1.6/; revision=12013
* Bug #247370 - Use a private PORTAGE_TMPDIR for --fetchonly mode in orderZac Medico2008-11-201-6/+12
| | | | | | | | | | | do avoid locking the normal build dir location (similar to bug #245231). The EbuildFetchPretend class is renamed to EbuildFetchonly and it's used for --fetchonly mode regardless of whether --pretend is enabled. Also, fix stdout/stderr flushing in JobStatusDisplay.displayMessage() and portage.spawn() to ensure output always occurs in the correct order when displaying pkg_nofetch() output in --fetchonly mode. (trunk r12006) svn path=/main/branches/2.1.6/; revision=12007
* Remove --fetchonly --resume time.sleep() call which is an artifact from theZac Medico2008-11-191-5/+0
| | | | | | previous parallel-fetch implementation. (trunk r12001) svn path=/main/branches/2.1.6/; revision=12002
* When warning about 'missing repo_name', also give the exact path where theZac Medico2008-11-181-15/+27
| | | | | | | | entry should exist, and explain that it should be a plain text file containing a unique name of the first line. This should give the users enough information to correct the problem without needing to seek help. (trunk r11993:11995) svn path=/main/branches/2.1.6/; revision=11996
* When a "slot conflict" occurs due to USE dependencies, display the usualZac Medico2008-11-161-1/+4
| | | | | | | slot conflict output together with the unsatisfied USE dependency output, since both might be useful. (trunk r11960) svn path=/main/branches/2.1.6/; revision=11961
* Fix logic inside PollScheduler._can_add_job() so that load average isZac Medico2008-11-151-1/+1
| | | | | | properly considered when a single job is running. (trunk r11958) svn path=/main/branches/2.1.6/; revision=11959
* Inside depgraph._add_pkg(), handle the case where multiple differentZac Medico2008-11-151-0/+14
| | | | | | | | | | instances of the same version (typically one installed and another not yet installed) have been pulled into the graph due to a USE dependency. The "slot collision" display is not helpful in a case like this, so display it as an unsatisfied dependency. Thanks to Peter Volkov <pva@g.o> for reporting. (trunk r11954) svn path=/main/branches/2.1.6/; revision=11955
* Bug #246853 - Redirect all ouput from depgraph.display_problems() to stderr,Zac Medico2008-11-151-3/+27
| | | | | | | except for unsatisfied dependency output which goes to stdout for parsing by programs such as autounmask. (trunk r11947:11949) svn path=/main/branches/2.1.6/; revision=11950
* Bug #236207 - Replace some hardcoded colors with color codes, and useZac Medico2008-11-151-23/+33
| | | | | | portage.output.EOutput() where appropriate. (trunk r11935) svn path=/main/branches/2.1.6/; revision=11936
* Bug #173284 - Update config update counting code in chk_updated_cfg_files()Zac Medico2008-11-151-3/+3
| | | | | | so hidden directories are ignored. (trunk r11933) svn path=/main/branches/2.1.6/; revision=11934
* For consistency with earlier portage-2.1.x releases, override SetArg.__str__()Zac Medico2008-11-141-0/+3
| | | | | | so that system and world sets don't show a leading @ character when displayed. svn path=/main/branches/2.1.6/; revision=11915
* Remove --search set matching support.Zac Medico2008-11-131-21/+2
| | | | svn path=/main/branches/2.1.6/; revision=11888
* Refer to portage-2.1.6_rc1 in comment about resume opts being stored as a dict.Zac Medico2008-11-131-1/+1
| | | | svn path=/main/branches/2.1.6/; revision=11887
* Update docs to not prefix system and world sets with the @ symbol.Zac Medico2008-11-131-1/+1
| | | | svn path=/main/branches/2.1.6/; revision=11885
* Remove preserve-libs support, it's API, and code that depends on it.Zac Medico2008-11-131-244/+0
| | | | svn path=/main/branches/2.1.6/; revision=11881
* Disable set expansion inside expand_set_arguments() and emulate existingZac Medico2008-11-131-144/+31
| | | | | | | portage-2.1.x behavior which treats system and world mutually exclusive actions. svn path=/main/branches/2.1.6/; revision=11880
* Remove sets protection code from unmerge().Zac Medico2008-11-131-88/+1
| | | | svn path=/main/branches/2.1.6/; revision=11879
* Update all portage.sets references to point to the new location, and hardcodeZac Medico2008-11-131-3/+3
| | | | | | the SetConfig constructor with a config for just system and world sets. svn path=/main/branches/2.1.6/; revision=11877
* Make post_emerge() exit early if it detects that the vdb state hasn't changed.v2.2_rc14Zac Medico2008-11-111-0/+9
| | | | | | | This works by comparing a hash of the COUNTER values for all packages in the vdb. svn path=/main/trunk/; revision=11859
* Allow --keep-going to continue in some cases when a runtime dependency hasZac Medico2008-11-111-2/+9
| | | | | | | failed to build or install. This involves pruning off the parts of the graph containing installed packages with unsatisfied dependencies. svn path=/main/trunk/; revision=11858
* Simplify code for cloning config instances inside Scheduler.merge().Zac Medico2008-11-111-1/+1
| | | | svn path=/main/trunk/; revision=11856
* Fix logic from previous commit in Scheduler._choose_pkg() to ensure thatZac Medico2008-11-111-1/+2
| | | | | | the --nodeps code is only triggered when --jobs > 1. svn path=/main/trunk/; revision=11848
* When in --nodeps mode, make Scheduler._set_digraph() discard the digraph andZac Medico2008-11-111-3/+4
| | | | | | make Scheduler._choose_pkg() always return the task at the front of the queue. svn path=/main/trunk/; revision=11847
* Inside depgraph.loadResumeCommand(), initialize self._scheduler_graph whenZac Medico2008-11-111-0/+1
| | | | | | | in --nodeps mode in order to avoid an AttributeError later when self.schedulerGraph() is called. svn path=/main/trunk/; revision=11846
* Fix --jobs parallel scheduling to ensure that temporary simultaneousZac Medico2008-11-101-39/+35
| | | | | | | | | installation of conflicting packages is avoided when appropriate (especially for !!atom blockers), but allowed in specific cases that require it. This is accomplished by reversing specific uninstall edges in the digraph, while possibly leaving some edges in there original state. svn path=/main/trunk/; revision=11845
* Revert r11839 since it's not necessarily desired to invert all uninstallZac Medico2008-11-101-4/+5
| | | | | | | edges. TODO: Invert only the specific edges that the depgraph has decided are necessary and allowed to be inverted. svn path=/main/trunk/; revision=11844
* Remove unused formatter code in show_invalid_depstring_notice().Zac Medico2008-11-091-3/+0
| | | | svn path=/main/trunk/; revision=11841
* Make show_invalid_depstring_notice() send output to stderr via writemsg_level().Zac Medico2008-11-091-12/+5
| | | | svn path=/main/trunk/; revision=11840
* Inside Scheduler._reverse_uninstall_edges(), iover all nodes rather than justZac Medico2008-11-091-1/+4
| | | | | | | | | | the merge list, because some uninstall nodes may not be in the merge list since they will be performed as part of an upgrade within a slot. This solves a problem with Scheduler._choose_pkg() not parallelizing in some cases when it should, due to an uninstall node not having it's edge reversed like it's supposed to. svn path=/main/trunk/; revision=11839
* Move initialization of Scheduler._background from the constructor to theZac Medico2008-11-091-13/+32
| | | | | | | | merge() method so that a potential InvalidDependString exception can be handled there, causing merge() to return unsuccessfully. This avoids having to raise an exception from the constructor. svn path=/main/trunk/; revision=11838
* Add PROPERTIES=interactive support in depgraph.display() and theZac Medico2008-11-091-0/+20
| | | | | | PackageCounters class. svn path=/main/trunk/; revision=11837
* Fix $ROOT handlink inside display_preserved_libs().Zac Medico2008-11-081-3/+3
| | | | svn path=/main/trunk/; revision=11829
* Handle CommandNotFound exceptions if the scanelf binary happens to be missing,Zac Medico2008-11-081-12/+22
| | | | | | and disable preserve-libs code in that case. svn path=/main/trunk/; revision=11827
* Remove the special package.keywords mask warning from bug #223447 sinceZac Medico2008-11-061-34/+4
| | | | | | | --depclean now uses the depgraph class and therefore should be more consistent with the update algorithm. svn path=/main/trunk/; revision=11815
* Bug #245661 - Never enable --complete-graph automatically since it's confusingZac Medico2008-11-061-14/+3
| | | | | | for users. svn path=/main/trunk/; revision=11814
* Synchronize portage update/restart logic in dblink.merge() andZac Medico2008-11-041-9/+3
| | | | | | | depgraph.display() so it's consistent with Scheduler._is_restart_necessary(). Thanks to Jeremy Olexa <darkside@g.o> for reporting. svn path=/main/trunk/; revision=11808
* Free some memory by clearing the LinkageMap cache when it's not needed.Zac Medico2008-11-041-0/+1
| | | | svn path=/main/trunk/; revision=11807
* Bug #245358 - For unsatisfied dependencies, display the parent nodes and theZac Medico2008-11-031-6/+18
| | | | | | argument that pulled them in. svn path=/main/trunk/; revision=11805
* Remove duplicate code inside EbuildFetchPretend._execute().Zac Medico2008-11-021-3/+0
| | | | svn path=/main/trunk/; revision=11801
* Remove locking from EbuildFetchPretend since it uses a private temp directoryZac Medico2008-11-021-5/+10
| | | | | | now (since bug #245231). svn path=/main/trunk/; revision=11800
* Bug #245231 - Make EbuildFetchPretend use a private temp directory in orderZac Medico2008-11-021-2/+18
| | | | | | to avoid locking the main one. svn path=/main/trunk/; revision=11796
* In Scheduler._calc_resume_list(), use self.pkgsettings instead of the theZac Medico2008-11-011-1/+1
| | | | | | temportary depgraph's. svn path=/main/trunk/; revision=11787
* Remove unused imports.Zac Medico2008-11-011-3/+3
| | | | svn path=/main/trunk/; revision=11774
* Use portage.const.PORTAGE_PACKAGE_ATOM where appropriate.Zac Medico2008-10-301-2/+4
| | | | svn path=/main/trunk/; revision=11758
* Bug #244975 - Don't allow the user to make portage unmerge itself sinceZac Medico2008-10-301-1/+24
| | | | | | there's never any valid reason to do so. svn path=/main/trunk/; revision=11756