Bug 701506 – distutils-r1.eclass: DISTUTILS_IN_SOURCE_BUILD + any-r1 API may result in 'cd' to missing BUILD_DIR (was: dev-python/jinja-2.10.3-r1 - cd: /var/tmp/portage/dev-python/jinja-2.10.3-r1/work/jinja-2.10.3-python3_8: No such file or directory)

• Status: RESOLVED FIXED
• Product: Gentoo Linux

[Description email200202 2019-11-30 00:41:17 UTC]

dev-python/jinja-2.10.3-r1 failed with message:

 * python3_8: running distutils-r1_run_phase python_compile_all
/var/tmp/portage/dev-python/jinja-2.10.3-r1/temp/environment: line 976: cd: /var/tmp/portage/dev-python/jinja-2.10.3-r1/work/jinja-2.10.3-python3_8: No such file or directory
 * ERROR: dev-python/jinja-2.10.3-r1::gentoo failed (compile phase):
 *   (no error message)
 * 


Reproducible: Always

Steps to Reproduce:
1. emerge --sync
2. emerge -auvDN world

Actual Results:  
dev-python/jinja-2.10.3-r1 failed to build

Expected Results:  
dev-python/jinja-2.10.3-r1 should not fail

[Comment 1 email200202 2019-11-30 00:42:35 UTC]

Created attachment 597936 [details]
build.log

[Comment 2 email200202 2019-11-30 00:43:41 UTC]

Created attachment 597938 [details]
emerge -pqv '=dev-python/jinja-2.10.3-r1::gentoo'

[Comment 3 email200202 2019-11-30 00:44:26 UTC]

Created attachment 597940 [details]
emerge --info '=dev-python/jinja-2.10.3-r1::gentoo'

[Comment 4 email200202 2019-12-01 12:42:45 UTC]

It compiles if you add python3_8 to PYTHON_TARGETS in make.conf.

[Comment 5 Michał Górny 2019-12-01 13:01:26 UTC]

I'll check this later today.

[Comment 6 Michał Górny 2019-12-01 13:03:37 UTC]

In the meantime, could somebody include eclass-debug.log?

[Comment 7 Michał Górny 2019-12-01 13:05:37 UTC]

Never mind, I see it's USE=-doc. I see the problem now.

[Comment 8 Michał Górny 2019-12-01 15:39:25 UTC]

Ok, so the problem is that the DISTUTILS_IN_SOURCE_BUILD API assumes each phase is run in BUILD_DIR (which normally is a copy of S).  However, any-r1 API assumes that the common phases may be run for impl that isn't being built at all, so the BUILD_DIR may not exist.  Normally, this isn't a problem since BUILD_DIR is not being used.  However, if you mix the two...

I'm still wondering what the correct solution to this should be.  We might just stay in S if BUILD_DIR doesn't exist.  We may also try creating it (copying S) when needed for any-r1 API.

[Comment 9 Larry the Git Cow 2019-12-01 20:19:47 UTC]

The bug has been closed via the following commit(s):

https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=f9b70d39ab2c06e476f3e8dc09d1d8f16e74e099

commit f9b70d39ab2c06e476f3e8dc09d1d8f16e74e099
Author:     Michał Górny <[e-mail]>
AuthorDate: 2019-12-01 16:12:55 +0000
Commit:     Michał Górny <[e-mail]>
CommitDate: 2019-12-01 20:19:34 +0000

    distutils-r1.eclass: Handle missing BUILD_DIR for any-r1 API
    
    When any-r1 API is used for *_all sub-phases, an implementation that
    was not selected by the user may be used.  In this case,
    DISTUTILS_IN_SOURCE_BUILD will not create the BUILD_DIR for this
    implementation; however, the phase runner will still try to enter it
    and fail.  Handle this gracefully by falling back to ${S}.
    
    Closes: https://bugs.gentoo.org/701506
    Signed-off-by: Michał Górny <[e-mail]>

 eclass/distutils-r1.eclass | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

[Comment 10 Michał Górny 2019-12-01 21:32:37 UTC]

*** Bug 701488 has been marked as a duplicate of this bug. ***