Re: GSoC: Porting Guix to Hurd week 3+4 report.

[ludo@gnu.org (Ludovic Courtès)]

Hi Manolis,

Thanks for the report!

Manolis Ragkousis <manolis837@gmail.com> skribis:

> 3) Found a circular dependency between glibc-hurd-headers and
> hurd-minimal. Resolved it
> and sent a patch to the list. (Ludovic please give it a look :-))

Will do shortly, sorry for the delay!

> 4) tarball-package in make-bootstrap.scm does not give the right name
> to the packages it produces.
> Changed tarball-package so now we can pass the target to it and as a
> result it will use the proper name.
>
> 5) gcc-4.7 passes "--with-native-system-header-dir=" which points to
> the wrong libc. According to my
> understanding this should point to the proper libc to be used in the
> target system. Am I right?
>
> 6) So the problem with %gcc-static is that libdecnumber: sets "dpd"
> while  libgcc: sets "no" and we get
>  a build failure because it can't find "no" in libdecnumber. Found a
> similar case here
> https://gcc.gnu.org/ml/gcc/2007-03/msg00941.html . Turned everything
> in make-bootstrap.scm into procedures
> so I am sure it evaluates to the right libc,

Note that all three points are related, AIUI.  That is, using procedures
in make-bootstrap.scm means that the right glibc package gets chosen,
which in turn means that #4 and #5 get fixed.

> patched libdecunmber's and libgcc's configure.ac so they run
> AC_CANONICAL_{BUILD,HOST,TARGET} and made sure with the repl that the
> right glibc is used.  And still can't find how to solve it. Any
> suggestions?

Does the issue reported in the 2007 message above still holds?  What
exactly was the error when cross-compiling libgcc?  The Hurd folks may
have seen it before.

Thanks for all the uneasy & frustrating work!

Ludo’.