Manolis Ragkousis <manolis...@gmail.com> skribis:
> The problem is that when trying to build %gcc-static from
> make-bootstrap.scm I get:
>
> make[2]: Entering directory
> '/tmp/nix-build-gcc-static-4.8.4.drv-0/build/i686-pc-gnu/libgcc'
> # If this is the top-level multilib, build all the other
> DEFINES='' HEADERS='' \
> ../../../gcc-4.8.4/libgcc/mkheader.sh > tmp-libgcc_tm.h
> make[2]: *** No rule to make target
> '../../../gcc-4.8.4/libgcc/../libdecnumber/no/decimal32.c', needed by
> 'decimal32.o'. Stop.
> make[2]: *** Waiting for unfinished jobs....
> # multilibs.
>
> This happens because glibc-for-bootstrap return the wrong glibc, which
> gcc-static is trying to use.
The problem in ‘make-bootstrap.scm’ is that %glibc-stripped and all the
%*-bootstrap-tarball variables are evaluated at the top level, thus
based on the default values of %current-system and
%current-target-system ("x86_64-linux" and #f here, whereas you would
like "i686-gnu" for the latter.)
One solution would be to turn all of these into procedures. That way,
%current-target-system would have the value you expect, and so
‘glibc-for-bootstrap’ & co. would use ‘glibc/hurd’.
Does that make sense?
Let me know how that works.
Thanks,
Ludo’.
