[bug#39480] [bug#47006] [WIP PATCH v2 2/2] gnu: Add zig.

[Liliana Marie Prikler]

Hi Sarah,

Am Samstag, den 11.09.2021, 12:24 -0700 schrieb Sarah Morgensen:
> All,
> 
> Apologies for the empty email earlier.  That shows me for trying to
> send from mobile!
Curse those mobile applications with their small buttons amirite? :P

> Liliana,
> 
> Liliana Prikler <liliana.prikler@gmail.com> writes:
> 
> > I've added a patch to use explicit search paths rather than
> > whatever Zig used
> > before and tried fixing some (syntactic) errors with the tests, but
> > was
> > unsuccesful, as there appear to be failing tests in the suite
> > itself.  Could
> > you have a look at the revised patch and check what flags you could
> > add to
> > the check phase to make it meaningful?
> > 
> > Btw. I haven't checked whether my cosmetic changes to #:configure-
> > flags break
> > things or not.  The end of the build phase puts a large amount of
> > stress onto
> > my system that I'd like to avoid at this hour.
> 
> I'm still working through the tests, but I did find one issue that
> has cropped up either from your patch or from the 0.7.1 -> 0.8.1
> upgrade.  This is from attempting to build tetris [0] (though I had
> to make a few syntax fixes, attached below, to build with 0.8.1):
W.r.t. the syntax fixes, that is probably an upstream issue or perhaps
an incompatibility introduced by zig itself.  (Maybe already from 0.7.1
to 0.8.0?)

> --8<---------------cut here---------------start------------->8---
> Zig attempted to find the path to native system libc headers by
> executing this command:
> cc -E -Wp,-v -xc /dev/null
> error: unable to create compilation: UnableToSpawnCCompiler
> --8<---------------cut here---------------end--------------->8---
It appears zig tries to execute @command{cc}.  Note, that this command
does not exist in Guix unless you install a symlink, so it will always
fail.  We might want to investigate the source of this error to check
whether spawning a C compiler is indeed the right choice for what
they're claiming to try or whether our hard coding already takes care
of that.

> No combination of ZIG_LIB_DIRS and ZIG_INCLUDE_DIRS seems to fix
> it.  Neither does --search-prefix.
> 
> If I set CC=gcc, it works fine.  But I think something changed such
> that it now has to fall back to this method of detection.  I have no
> idea why.
Thanks for the info.  I'll let you know once I find out more.