To avoid variable name clash with ports' DISTDIR, rename DISTDIR used for pkgbase to PBDISTDIR
Diff Detail
- Lint
Lint Skipped - Unit
Tests Skipped
Event Timeline
@kevans pointed out that:
02:45 < kevans91> otis: your DISTDIR problem is a tricky one
02:46 < kevans91> otis: PBDISTDIR isn't technically correct name as it's used for much more than pkgbase... but the conflict isn't a good one
02:47 < kevans91> I suspect we should expand it to DISTRIBDIR or something instead, but we need to comb the rest of the tree and figure out the impact
02:47 < kevans91> At least release/ will need to change
02:48 < kevans91> Or RELDISTDIR
02:49 < kevans91> share/man/man7/build.7 also needs to reflect the change
I'll do further investigation how this could be done in an ellegant way.
- Rename DISTDIR to RELDISTDIR
- Reflect this change also in build(7)
- Make an entry in UPDATING
At first glance this looks ok.
I'll do some test build + release.sh build this weekend to confirm that everything is still working.
Thanks
I think this is almost complete; there's just two occurrences in share/mk that are also relevant and should be replaced, one in a comment in bsd.own.mk and one in distribute target of bsd.subdir.mk.
Also adding re@, because we should get them to confirm if DISTDIR is used in a base-context in thermite for release-building. I would suspect not, though.
Yes, can we please get this committed ASAP? I just spent some time debugging this problem. (Other solution would be for base not to read /etc/make.conf, but I have a feeling this commit is more likely.)
| Makefile.inc1 | ||
|---|---|---|
| 926 | Given we're renaming the variable and touching every line, I wonder if we can't move the / into the value and drop this METALOG:= line. | |
I also just ran into this after putting DISTDIR=/var/cache/mydistfiles for Ports into my $HOME/.profile and took an hour of debugging. This patch doesn't apply anymore as it's been years, but the fix is still needed.
Either ports shouldn't be leaking DISTDIR into the base environment or the base build system here should be sanitising its environment. Renaming it just papers over the underlying issue.
Also, this is hardly a new variable. Why is it only now that it's such an urgent issue to fix?
| Makefile.inc1 | ||
|---|---|---|
| 926 | What do you mean? This canonicalises (in terms of duplicate /s) whatever METALOG is set, regardless of whether it comes from the line prior or set by the environment / make arguments. | |