Re: gengenrtl needs to link with libintl

[DJ Delorie <dj at redhat dot com>]

> The above is reasonable but the current reality is that obstack.c is
> compiled for gcc with the macro GENERATOR_FILE defined.  I submitted
> a patch to check for GENERATOR_FILE in obstack.c a number of days ago,
> <http://gcc.gnu.org/ml/gcc-patches/2001-10/msg01695.html>, but there
> has been no specific comment or review.

In general, I oppose changes to libiberty that:

1. Are gcc-specific, as libiberty is used in far more than just gcc.

2. Make libiberty's files differ from the master files (glibc in this
   case) (although those files already differ, making them differ more
   is still bad).

> In my opinion, when a problem develops causing bootstrap failure on one
> or more platforms, the issue needs to be addressed with more urgency
> than bugs that have a more limited impact.  I'd prefer to have a pragmatic
> solution in place while the issues are debated by the maintainers of
> libiberty and the gcc build process.  This could either be a temporary
> fix or reversal of the patch which caused the failure.  We seem to be
> stalled in this case because another patch to resolve the issue didn't
> receive comment.

The generic agreed-upon course to take when these things happen is to
revert whatever caused the problem to manifest.  Long term, IMHO the
perferred solution is to have a build-libiberty so that this gcc
hackery can be removed.

http://gcc.gnu.org/ml/gcc-patches/2001-07/msg00608.html

And please feel free to bug me privately if I don't respond to a
posted patch in a day or two.