Nikita Karetnikov <nik...@karetnikov.org> skribis:
>> The solution is to patch Python to refer to our ‘sh’ instead of /bin/sh
>> (as is done for Guile’s ice-9/popen.scm).
>
>> Can you do that?
>
> I changed 'arguments', but 'gobject-introspection' failed with the same
> error.
>
> python.scm:
>
> (arguments `(#:tests? #f ; XXX: some tests fail
> #:phases (alist-cons-before
> 'configure 'pre-configure
> (lambda* (#:key inputs #:allow-other-keys)
> (let ((bash (assoc-ref inputs "bash")))
> (substitute* "Lib/subprocess.py"
> (("/bin/sh")
> (string-append bash "/bin/bash")))))
> %standard-phases)))
Good. You can already commit that one (but please make it
(string-append bash "/bin/sh") rather.)
> subprocess.py:
>
> if shell:
> args =
> ["/nix/store/4mg8b8vvmava68y64qmm70gqfnhhjzmx-bash-4.2/bin/bash", "-c"] + args
> if executable:
> args[0] = executable
Are you looking at the subprocess.py that’s actually mentioned in the
gobject-introspection backtrace?
If yes, then it likely means that it’s the caller–i.e., some .py file in
gobject-introspection–that’s passing /bin/sh somewhere. Can you check
the files mentioned in the backtrace?
HTH,
Ludo’.
