bug#29403: [PATCH] gnu: gcl: Update snapshot.

[Kei Kebreau <kkebreau@posteo.net>]

[-- Attachment #1: Type: text/plain, Size: 2205 bytes --]

ludo@gnu.org (Ludovic Courtès) writes:

> Kei Kebreau <kkebreau@posteo.net> skribis:
>
>>> Kei Kebreau <kkebreau@posteo.net> skribis:
>>>
>>>> * gnu/packages/lisp.scm (gcl): Update to 2.6.12-1.5956140.
>>>> [arguments]: Remove CFLAGS from make-flags; adjust pre-conf phase.
>>>> [native-inputs]: Add which.
>>>
>>> Could you include a rationale for using a development snapshot?
>>> Normally we’d stick to upstream releases, unless there’s a good reason.
>>>
>>> Thanks,
>>> Ludo’.
>>
>> The short story is that there is an issue with GCL's garbage collection
>> when it gets built on hydra.gnu.org.
>
> Do you mean that substitutes disappeared?  (I just tried and got
> substitutes from berlin.guixsd.org.)
>

Sorry for not being clearer. I meant that the stratified garbage
collection that GCL uses seems to have a non-reproducible memory
protection issue right now.

>> In the process of investigating this issue, I noticed that Camm
>> Maguire, GCL maintainer, uses these development snapshots in the
>> package he oversees for Debian. This specific snapshot is the latest
>> one and is shipped in Debian unstable, but even stable uses a fairly
>> recent GCL development snapshot. These snapshots are mainly small bug
>> fixes and optimizations from the look of the GCL commit history.
>
> OK, that certainly justifies this patch.
>
> Nevertheless, we should also kindly invite them to publish these as
> formal releases rather than giving Debian a special treatment.
>

I'll send a message to the GCL mailing list.

>> If you think this update is permissible, how does the following commit
>> message read:
>>
>>     * gnu/packages/lisp.scm (gcl): Update to 2.6.12-1.5956140.
>>     [arguments]: Remove CFLAGS from make-flags; adjust pre-conf phase.
>>     [native-inputs]: Add which.
>>
>>     This update includes small upstream bug fixes and optimizations.
>
> Sounds good, yes.
>
> Thanks for explaining!
>
> Ludo’.

Pushed to master as dd0134fcb707452e1c343d66af6088c0be38a285. I also
managed to avoid adding "which" as a build dependency by substituting
paths to gcc, ld and objdump in GCL's top loop. Thanks for the review!

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 832 bytes --]