[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[bug#31952] [PATCH] gnu: git: Update to 2.18.0.

From: Ludovic Courtès
Subject: [bug#31952] [PATCH] gnu: git: Update to 2.18.0.
Date: Sun, 24 Jun 2018 22:01:20 +0200
User-agent: Gnus/5.13 (Gnus v5.13) Emacs/26.1 (gnu/linux)


Marius Bakke <address@hidden> skribis:

> address@hidden (Ludovic Courtès) writes:
>> Hi Marius!
>> Marius Bakke <address@hidden> skribis:
>>> * gnu/packages/version-control.scm (git): Update to 2.18.0.
>>> [native-inputs]: Add BASH.
>>> [arguments]: In #:make-flags, define SHELL_PATH and TEST_SHELL_PATH.
>>> Add #:disallowed-references.  Drop /bin/sh substitution from Makefile.  Add
>>> phase to prevent BASH from ending up in PATH.  Delete 'patch-shebangs' 
>>> phase.
>> [...]
>>> +        ;; FIXME: This phase picks up the wrong bash when patching 
>>> shebangs.
>>> +        (delete 'patch-shebangs)
>> Do the installed scripts still have the right shebang in spite of this?
> I haven't yet compared the results with and without this phase, but
> "normal" usage (including "send-email") works at least.

OK.  I suppose you can simply grep for “^#!” and make sure there’s
nothing suspicious like /usr/bin/something.

>> Removing this phase altogether sounds a bit risky.  Another option would
>> have been to replace it with one that moves the “right” Bash to the
>> front of PATH and then calls the original ‘patch-shebangs’ phase.
> The problem is that 'patch-shebangs' does not use PATH, but instead
> iterates over inputs directly.  It's supposed to prefer 'inputs' to
> 'native-inputs' (according to a comment), yet in this case it picks the
> native "full" bash rather than bash-minimal.


> If you read closely, you'll notice that 'bash-for-tests' is not in PATH
> at all.  'patch-source-shebangs' and other things that use (which "sh")
> works okay due to that.

It may be that moving “bash-for-tests” to ‘inputs’ actually solves the
problem (and we don’t have to worry about cross-compilation since “make
check” does nothing when cross-compiling) and we can keep the
‘patch-shebangs’ phase.

But yeah, if this patch works, we can go for it.

Thank you!


reply via email to

[Prev in Thread] Current Thread [Next in Thread]