Following the same as https://reviews.freebsd.org/D53053
in the src tree.
Details
- Reviewers
lwhsu bapt - Group Reviewers
portmgr - Commits
- R11:53c1ad8c4add: bsd.port.mk: run pkg-create with MAKE_JOB_NUMBER threads
Diff Detail
- Repository
- R11 FreeBSD ports repository
- Lint
Lint Not Applicable - Unit
Tests Not Applicable
Event Timeline
(Is commenting on a closed review the Right Thing[tm] to do? Or should I be creating a separate one so as not to confuse things?)
I have a concern here, beyond the reproducibility/ratio/etc. comments that are being touched on in replies to the commit message.
Tying this explicitly to MAKE_JOBS_NUMBER means that when MAKE_JOBS_UNSAFE is set, packaging only gets one thread. Usually, one sets that latter variable when the Makefile has parallelism/ordering issues, and so having that also affect how packaging works is probably going to be unexpected behaviour.
Could we maybe introduce a variable for this purpose? I'm (attempting to) attach a patch to this comment (so as not to affect the main diff here) that adds in a PKG_CREATE_THREADS_NUMBER variable and sets it using some of the same logic for MAKE_JOBS_NUMBER.
Plus, given that there are some concerns being brought up on the list, this would be a way for people to continue having build parallelism, but avoid any unexpected side effects of multi-threaded packaging.
edit: Fixing comment that was in the completely wrong place.
{F146697295}