Hello,

ludo@gnu.org (Ludovic Courtès) writes:

> Efraim Flashner <efraim@flashner.co.il> skribis:
>
>> I believe this would cause gcc-5 to be rebuilt, which would cause a
>> world rebuild, so for the moment we also need a native-inputs entry for
>> gcc-5 making sure that nothing changes, ie: that we're only using
>> texinfo there, preferably with a note that it should be changed in the
>> future.  Also, you should take a look at gnu/packages/commencement and
>> make sure that we're not adding perl as a native-input for gcc-boot0 or
>> for libstdc++, wherever that happens to live (I don't remember off
>> hand).
>
> I think this should go to ‘core-updates’.
>
>> I assume it is OK for us to build the man pages for gcc-final (if it
>> happens) since this is the gcc used for actually building packages, and
>> I believe the one pulled in for gcc-toolchain.
>
> More precisely, building man pages for ‘gcc-boot0’ is pointless and
> possibly adds Perl on the graph for no good reason.  So if we can
> arrange to just have man pages in ‘gcc-final’, we’re good.
>
> Efraim, could you check whether that works for ‘core-updates’?
>
> Thanks,
> Ludo’.

I believe this is what is accomplished by the attached patch v3. Thanks
to Efraim for the points raised. It turned out that the stack overflow I
was getting attempting to test the previous version was caused by a
circular dependency that was introduced by gcc-boot0 inheriting
perl. This is now fixed.

I spent the last couple hours rebuilding gcc but it failed due to lack
of space on my side... Retrying now but it'll take a few hours before I
can confirm that it works as intended.

Thank you both!