summaryrefslogtreecommitdiffstats
path: root/bin/isolated-functions.sh
Commit message (Collapse)AuthorAgeFilesLines
* Bug #219286 - Whitelist INSTALL_MASK so that it works properly for binaryZac Medico2008-05-051-0/+3
| | | | | | | | packages. Also, filter INSTALL_MASK when saving or loading environment.bz2. This also applies to PKG_INSTALL_MASK and DOC_SYMLINKS_DIR, so add them too. svn path=/main/trunk/; revision=10195
* Handle issues with newlines in elog messages that can trigger an unhandledZac Medico2008-04-091-1/+1
| | | | | | | | | ValueError to be raised from a split() call inside collect_ebuild_messages(): * Use \0 to delimit messages, so that that elog messages containing newlines are handled correctly. * Handle a potential ValueError when splitting the message type. svn path=/main/trunk/; revision=9763
* Fix BASH_ARGC and BASH_ARGV offsets some more.Zac Medico2008-04-061-1/+1
| | | | svn path=/main/trunk/; revision=9734
* In dump_trace(), fix BASH_ARGV and BASH_ARGC offsets so that they are alwaysZac Medico2008-04-061-3/+9
| | | | | | correct wrt eachother. Thanks to Betelgeuse for reporting. svn path=/main/trunk/; revision=9733
* preserve order of ebuild messages even between different message types (bug ↵Marius Mauch2008-04-061-1/+1
| | | | | | #197905) svn path=/main/trunk/; revision=9726
* Bug #214619 - Add support for a PORTAGE_BINPKG_TAR_OPTS variable thatZac Medico2008-03-281-1/+1
| | | | | | | allows the user to specify tar command options for binary package creation. svn path=/main/trunk/; revision=9528
* Remove the killparent() function and associated SIGINT trap since thisZac Medico2008-02-101-1/+1
| | | | | | | should already be handled on the python side and it won't work with dropped privileges anyway. svn path=/main/trunk/; revision=9327
* Bug #155993 - Update environment.bz2 after postinst in caseZac Medico2008-02-071-1/+2
| | | | | | | installation phases need to pass some variables to uninstallation phases. svn path=/main/trunk/; revision=9281
* - Remove ebuild.autotools repoman check since it produces lotsZac Medico2008-01-301-1/+2
| | | | | | | | | of false positives (bug #207104) and its hard to parse bash - Use QA_INTERCEPTORS in ebuild.sh to implement a runtime 'QA Notice' when autotools is called directly svn path=/main/trunk/; revision=9235
* We need to implement diefunc() since environment.bz2 files containZac Medico2008-01-211-1/+7
| | | | | | calls to it (due to alias expansion). svn path=/main/trunk/; revision=9217
* Replace the current diefunc() and die alias with a single die() function.Zac Medico2008-01-201-6/+3
| | | | | | | | | The previous arguments to diefunc() weren't used anyway. The result is that the die message looks better because it doesn't show the alias and all of it's arguments which are totally irrelevant. (branches/prefix r9212) svn path=/main/trunk/; revision=9216
* Make filter_readonly_variables() filter PATH in any case sinceZac Medico2008-01-051-1/+1
| | | | | | we never want to save of reload it. svn path=/main/trunk/; revision=9147
* Send eerror and ewarn messages to stderr. This is important inZac Medico2008-01-041-2/+2
| | | | | | | | case somebody calls portageq inside a global scope command substitution. The output needs to go to stderr so that the die message from the subshell is visible. svn path=/main/trunk/; revision=9140
* As suggested by grobian, treat Darwin just like USERLAND=gnuZac Medico2007-12-291-4/+1
| | | | | | since they seem to be switching commands to gnu anyway. svn path=/main/trunk/; revision=9118
* Fix broken Darwin USERLAND detection.Zac Medico2007-12-291-0/+3
| | | | svn path=/main/trunk/; revision=9116
* Simplify the userland detection code in portage.data and addZac Medico2007-12-291-0/+11
| | | | | | a bash version of it to isolated-functions.sh. svn path=/main/trunk/; revision=9100
* Move XARGS initialization from portage.data to isolated-functions.sh.Zac Medico2007-12-291-0/+11
| | | | svn path=/main/trunk/; revision=9096
* Fixes for logic related to FEATURES=test to USE=test mapping:Zac Medico2007-12-281-1/+1
| | | | | | | | | | | | | | | * Add EBUILD_FORCE_TEST to the environment whitelist and filter it from the saved environment in save_ebuild_env(). * Tweak logic inside the ebuild command and config.regenerate() so that EBUILD_FORCE_TEST works even in odd cases like when USE=test is masked. * Only make FEATURES=test map to USE=test when "test" is actually in IUSE. * Remove USE=test from the set of implicit IUSE so that useq() calls in ebuild.sh properly generate a QA Notice when "test" is missing from IUSE. svn path=/main/trunk/; revision=9064
* Make sure that PORTAGE_SANDBOX_* variables can never be setZac Medico2007-12-131-1/+1
| | | | | | by "${T}"/environment. svn path=/main/trunk/; revision=8911
* Allow FEATURES to persist between phases since the user mightZac Medico2007-12-131-1/+1
| | | | | | | | | | | want to modify it via bashrc to enable things like splitdebug and installsources for specific packages. They should be able to modify it in pre_pkg_setup() and have it persist all the way through the install phase. However, if FEATURES exist inside environment.bz2 then they should be overridden by current settings. svn path=/main/trunk/; revision=8909
* * whitelist CCACHE_* and DISTCC_* variables in config.environ()Zac Medico2007-12-031-0/+3
| | | | | | * unset CCACHE_* and DISTCC_* variables in save_ebuild_env() svn path=/main/trunk/; revision=8818
* For compatibility with cvs.eclass, do not unset CVS_RSH inZac Medico2007-12-021-1/+1
| | | | | | | save_ebuild_env() since that variable may be initialized in global scope. svn path=/main/trunk/; revision=8816
* * whitelist some misc variables from the calling environmentZac Medico2007-12-011-0/+4
| | | | | | | | * blacklist the same variables in save_ebuild_env() so that the latest values from the calling environment always override those from the ebuild environment svn path=/main/trunk/; revision=8796
* When die has been called and ${T}/environment does not exist,Zac Medico2007-12-011-1/+8
| | | | | | | dump the current environment to ${T}/die.env in case it helps for debugging. svn path=/main/trunk/; revision=8783
* * whitelist PREROOTPATH in config.environ()Zac Medico2007-12-011-1/+2
| | | | | | * filter PREROOTPATH in save_ebuild_env() svn path=/main/trunk/; revision=8781
* Bug #200863 - Don't filter ${KV} since kernel-2.eclass uses thatZac Medico2007-12-011-1/+1
| | | | | | variable and we don't want to interfere. svn path=/main/trunk/; revision=8770
* Bug #200775 - Save more portage generated environmentZac Medico2007-11-301-1/+1
| | | | | | | variables that are in environment.bz2 so that they are available for use by tools such as epm. svn path=/main/trunk/; revision=8765
* Move environment.bz2 extraction from ebuild.sh to doebuild() onZac Medico2007-11-291-1/+1
| | | | | | | | | | | | the python side. The python will be able to use it's awareness of the ${T}/environment to decide what type of ebuild environment should be generated. For example, if the ebuild environment should be able to unset variables that have been inherited from the calling environment, the existence of ${T}/environment will indicate that the ebuild environment should be isolated from the calling environment. svn path=/main/trunk/; revision=8753
* Check if ${EBUILD_EXIT_STATUS_FILE} is empty beforeZac Medico2007-11-261-1/+2
| | | | | | attempting to use it. svn path=/main/trunk/; revision=8686
* Bug #200313 - Detect and report when an ebuild phaseZac Medico2007-11-261-1/+4
| | | | | | | | | | | | | | | | exits unexpectedly. This is type of behavior is known to be triggered by things such as failed variable assignments (bug #190128) or bad substitution errors (bug #200313). We use a EBUILD_EXIT_STATUS_FILE environment variable to specify a file that the shell code is supposed to create when it exits in a normal manner. If the file does not get created like it's supposed to be then we can conclude that the shell has exited in some unexpected way. svn path=/main/trunk/; revision=8682
* Make diefunc() show the path of ${T}/environment since it's usuallyZac Medico2007-11-261-0/+2
| | | | | | in the stack trace and the user is likely to be interested in it. svn path=/main/trunk/; revision=8678
* Filter more misc environment variables.Zac Medico2007-11-261-2/+2
| | | | svn path=/main/trunk/; revision=8676
* * Move filtering of a bunch of portage config variables fromZac Medico2007-11-261-21/+10
| | | | | | | | save_ebuild_env() to config.environ(). * Add EMERGE_FROM to the READONLY_PORTAGE_VARS list. svn path=/main/trunk/; revision=8670
* Filter _RC_GET_KV_CACHE from the saved environment.Zac Medico2007-11-251-1/+1
| | | | svn path=/main/trunk/; revision=8666
* Make preprocess_ebuild_env() reload a known good version ofZac Medico2007-11-251-0/+68
| | | | | | | save_ebuild_env() in order to avoid making an assumption that it hasn't been overridden. svn path=/main/trunk/; revision=8663
* The qa_call() function and anything before it are portage internalsZac Medico2007-11-251-1/+11
| | | | | | | that the user will not be interested in. Therefore, the stack trace should only show calls that come after qa_call(). svn path=/main/trunk/; revision=8653
* Fix the diefunc overlay message so that it doesn't triggerZac Medico2007-11-211-0/+1
| | | | | | falsely in prerm or postrm. svn path=/main/trunk/; revision=8578
* Make diefunc() clean up QA_INTERCEPTORS in case it isZac Medico2007-11-111-1/+6
| | | | | | called from inherit() since sed called below. svn path=/main/trunk/; revision=8490
* Fix diefunc() so that it doesn't report that the ebuild is fromZac Medico2007-10-311-1/+2
| | | | | | an overlay when EMERGE_FROM == binary. svn path=/main/trunk/; revision=8349
* Fix quoting as reported by Drac in bug 194169Alec Warner2007-09-291-1/+1
| | | | svn path=/main/trunk/; revision=7878
* with help from Donnie and Ulrich, use sed to extract the statement that led ↵Mike Frysinger2007-09-271-2/+24
| | | | | | to the call to die so we can print it out in our error message svn path=/main/trunk/; revision=7852
* align the filename and line numbers in the output of the stack dump so that ↵Mike Frysinger2007-09-261-13/+21
| | | | | | it is much easier to read svn path=/main/trunk/; revision=7842
* someone seriously screwed the pooch on whitespacing here so clean it all upMike Frysinger2007-09-261-62/+63
| | | | svn path=/main/trunk/; revision=7839
* Fix quoting.Zac Medico2007-06-271-1/+1
| | | | svn path=/main/trunk/; revision=7057
* Format $EBUILD_OVERLAY_ECLASSES output more like it was before.Zac Medico2007-06-271-1/+5
| | | | svn path=/main/trunk/; revision=7055
* Format the callstack indentation more like it used to be.Zac Medico2007-06-271-2/+2
| | | | svn path=/main/trunk/; revision=7050
* Send the ebuild die message through eerror and make sure that elog_process() ↵Zac Medico2007-06-271-17/+18
| | | | | | gets called when an ebuild dies. This makes the elog echo module display the die message last in case other elog messages push the original die message off of the top of the screen. svn path=/main/trunk/; revision=7047
* Fix eend so that it works properly on a standard BSD console. Thanks to ↵Zac Medico2007-05-021-0/+2
| | | | | | UberLord. svn path=/main/trunk/; revision=6467
* die requires diefunc and dump_trace too; duh; also move extdebug into ↵Alec Warner2007-02-081-0/+31
| | | | | | isolated-functions.sh svn path=/main/trunk/; revision=5930
* Only echo PORTAGE_LOG_FILE if it's actually set (may not be if sesandbox is ↵Zac Medico2007-02-071-1/+2
| | | | | | enabled). svn path=/main/trunk/; revision=5912