bug#22354: Test failure when running distcheck from out-of-tree build

[Ludovic Courtès]

address@hidden (Taylan Ulrich "Bayırlı/Kammer") skribis:

> address@hidden (Ludovic Courtès) writes:
>
>> address@hidden (Taylan Ulrich "Bayırlı/Kammer") skribis:
>>
>>> If one runs distcheck from within the build directory of an out-of-tree
>>> build (perhaps a strange combination), the guix-environment.sh test
>>> fails with the following log snippet.
>>
>> [...]
>>
>>> ./configure: ./config.status: 
>>> /home/taylan/src/guix/build/guix-0.9.1/_build/sub/test-tmp/store/d41iyl2gyk0r:
>>>  bad interpreter: No such file or directory
>>
>> I think you’re hitting the shebang limit (127 chars) that prevents you
>> from running test from that build tree.  ./configure emits a warning in
>> this case, which you should see in ‘config.log’.
>>
>> So, not a bug, but it would have been nicer if ‘make check’ had failed
>> directly.  Commit bb25130 does that.
>>
>> Though now that I look again at the error above, I see that we’re below
>> the 127 char limit.  Weird.
>
> Hmm, that commit didn't fix the issue.

You mean it doesn’t error out directly upon ‘make check’?  Could you
run “grep length …/_build/config.log”?

> I'm also puzzled by the cut-off of the shebang string at 77 characters
> in the error message, so I looked into it a bit, but nothing strange
> here from what I can tell: the ./configure of make-4.1 is called with
> SHELL and CONFIG_SHELL set to
>
> /home/taylan/src/guix/build/guix-0.9.1/_build/sub/test-tmp/store/d41iyl2gyk0rwd9z2ddgph9nyg48f3p0-bootstrap-binaries-0/bin/bash
>
> which is exactly 127 characters, but add #! and you have 129.  This ends
> up being the shebang of config.status, which is why it fails while
> trying to call that.

So the diagnostic was correct.

> Only the error message is broken.

Would be interesting to see which is the piece of software that has an
even lower limit here.

Ludo’.