[bug#46725] [PATCH] gnu: guile-lib: Fix cross compilation.

[Maxim Cournoyer]

Hi Chris,

Christopher Baines <mail@cbaines.net> writes:

> Ludovic Courtès <ludo@gnu.org> writes:
>
>> Hi,
>>
>> Maxim Cournoyer <maxim.cournoyer@gmail.com> skribis:
>>
>>> Ludovic Courtès <ludo@gnu.org> writes:
>>>
>>>> Hi!
>>>>
>>>> Christopher Baines <mail@cbaines.net> skribis:
>>>>
>>>>> These changes were sent upstream as
>>>>> https://lists.gnu.org/archive/html/guile-devel/2021-02/msg00004.html
>>>>>
>>>>> Without this change, the .go files are built for the host architecture, 
>>>>> rather
>>>>> than the target. I noticed this when cross building the
>>>>> guix-build-coordinator (for which guile-lib is an input) to the Hurd.
>>>>>
>>>>> * gnu/packages/guile-xyz.scm (guile-lib)[arguments]: Add
>>>>> 'patch-for-cross-compilation phase.
>>>>> [native-inputs]: Add autoconf, automake and gettext.
>>>>> (guile2.0-lib): Adjust to use alist-replace.
>>>>> (guile2.2-lib): Adjust to use alist-replace.
>>>>
>>>> [...]
>>>>
>>>>> +             (substitute* "configure.ac"
>>>>> +               (("GUILE_FLAGS")
>>>>> +                "GUILE_FLAGS
>>>>> +if test \"$cross_compiling\" != no; then
>>>>> +   GUILE_TARGET=\"--target=$host_alias\"
>>>>> +   AC_SUBST([GUILE_TARGET])
>>>>> +fi
>>>>
>>>> You could even set ‘GUILE_TARGET’ unconditionally.
>>>
>>> It seems a conditional is needed, because info
>>> '(autoconf)Canonicalizing' says:
>>>
>>>     The variables ‘build_alias’, ‘host_alias’, and ‘target_alias’ are
>>>     always exactly the arguments of ‘--build’, ‘--host’, and ‘--target’; in
>>>     particular, they are left empty if the user did not use them, even if
>>>     the corresponding ‘AC_CANONICAL’ macro was run.
>>>
>>> I.e. without the condition '--target=' could be passed to guild, which
>>> probably wouldn't work.
>>
>> Ah true.  In Guile-{Gcrypt,zlib,…}, ‘--target’ is passed conditionally.
>> Apologies for the confusion!
>>
>>> So it seems that nowadays the build, host, and target would take default
>>> values as guessed by config.guess, and that the earlier section of the
>>> manual should be updated?
>>
>> Dunno; but in hindsight, what Chris did seemed right.
>
> I don't really get the change I'm making, but I'm confident it has the
> desired effect, I just copied it other Guile libraries. Anyway, I think
> this is still useful, so I've gone ahead and pushed as
> a0b24d5f8c5813329986b1bd690ed50b5308cc78.
>
> If there are better ways to get the cross compiling to work, then that
> can be put in place upstream, and then these changes removed.
>
> Thanks,
>
> Chris

FYI, I have sent to the Guile patches tracker commits (keeping original
authorship) of all the Guix-specific changes to guile-lib; the version
of your patch is slightly modified to test the value about to be used,
but otherwise it behaves the same.

Maxim