Leo Famulari <l...@famulari.name> writes:

> On Sat, Sep 24, 2016 at 01:13:33AM -0400, Kei Kebreau wrote:
>> Leo Famulari <l...@famulari.name> writes:
>> > I was getting failures reminiscent of an ABI break. But, I can't
>> > reproduce it now.
>> >
>> > Kei, maybe you should try applying the patches to a fresh Git checkout?
>> > If they work like that, then I think it's okay to push, assuming the
>> > qscintilla URL is fixed.
>> 
>> No luck...
>
> Hm, how about if you apply the first patch, `make`, then apply the
> second patch, and `make` again? Or some dance like that. Maybe deleting
> one of qt.go or maths.go in between. Something like that worked for me
> but I don't remember all the steps.
>
> Of course that is not acceptable for code pushed to the master branch,
> but if you figure out how to reproduce a "successful" case, it will make
> this easier to debug.

I found a lead on our problem!

Applying the qscintilla patch and building it with `guix build` works
fine.

Applying the qscintilla-qt4 patch (not even the octave one) and building
octave with `guix build` gives me the following:

  In unknown file:
     ?: 0 [memoize-variable-access! #<memoized fftw> #<directory # 51eaf30>]

  ERROR: In procedure memoize-variable-access!:
  ERROR: Unbound variable: fftw

I doubt that anything is wrong with the fftw package, because I'm only
given the error when I define qscintilla-qt4 (public or not). Should I
be doing something like what is done with the curl package,
i.e. defining the replacement variable? I'm kind of confused by this situation.

Attachment: signature.asc
Description: PGP signature

Reply via email to