aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorFlorian Schmaus <flow@gentoo.org>2024-04-02 22:28:30 +0200
committerMike Gilbert <floppym@gentoo.org>2025-02-01 12:13:46 -0500
commitc75b61b523457f8798c66afe4bcdc1708408ea00 (patch)
treea8cf752645117f75b652df8fd8e6da361a74ab90 /LICENSE
parentgetname: Handle None remotepkgs (diff)
downloadportage-master.tar.gz
portage-master.tar.bz2
portage-master.zip
Do not export PMS variables in EAPI 9 (or if FEATURES=-export-pms-vars)HEADmaster
Instead of passing PMS variables as part of the process environment, we pass them via a file that is later sourced by the ebuild's bash. Since, for example, A is usually the greatest contributor to the process environment, removing it from the process environment significantly avoids running into MAX_ARG_STRLEN when spawning a new child process. This means that A and other PMS variables are no longer exported in the ebuild and hence unavaiable to child processes. However, A is mostly used as part of the default_src_unpack function and there A does not need to be exported. This started as a change that only unexported A, but was later extended to most PMS variables as suggested by ulm in https://bugs.gentoo.org/721088#c23. Thanks to Zac Medico for helpful input on this change, and to Eli Schwartz for suggesting that (bash) helpers should simply source the environment file introduced by this change. Closes: https://bugs.gentoo.org/721088 Signed-off-by: Florian Schmaus <flow@gentoo.org> Closes: https://github.com/gentoo/portage/pull/1407 Signed-off-by: Mike Gilbert <floppym@gentoo.org>
Diffstat (limited to 'LICENSE')
0 files changed, 0 insertions, 0 deletions