This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: One version number, take three


Michael Matz <matz@suse.de> writes:

> Something is fishy with the patch.  The problem seems to be, that
> the top-level Makefile also tries to set gcc_version, or better the
> configure tries to set it from the version_trigger (which now is
> empty) and substitutes this into the toplevel Makefile.  So for me
> the Makefile now contains:
>
> gcc_version = const
>
> (I don't know where the word 'const' comes from)
>
> The problem with this is, that this is used to set libsubdir, also
> in the toplevel Makefile.  And that one is passed down to the
> Makefile in gcc.  During installation this directory is used to
> install things like crtbegin.o into.  So for me when calling make
> install from the toplevel, crtbegin.o is installed into
> prefix/lib/gcc/x86_64-unknown-linux-gnu/const/ instead of
> prefix/lib/gcc/x86_64-unknown-linux-gnu/4.1.0/ where it is
> (correctly) searched.
>
> So, although the Makefile in gcc/ is correct, and would use the
> correct libsubdir, the toplevel Makefile makes this break.

Foolish me, I never thought to check the toplevel Makefile for such
things.  gcc_version_trigger and libsubdir references turn out to be
sprinkled all over the target libraries.  I am working on a fix now.

zw


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]