https://gcc.gnu.org/bugzilla/show_bug.cgi?id=123976

--- Comment #116 from Peter Dyballa <Peter_Dyballa at web dot de> ---
The last few hours I was trying to understand what the patches from
patch-darwin-gcc-16.1.0.diff might be doing. I thought it would be easy,
because I saw a few .TGT files – but they're in the testsuite! So it took some
time – and I think this patch file is not the culprit. There are strange .c,
.h, .cc files that seem to have an impact on the build process, but I do not
understand how.

IMO it's at least one missing make rule. 'make -p' prints kind of a text db
with all the existing rules. And the missing three symlinks are mentioned
there. In the debug output from 'make -d' is no sign of them. They're out,
away, got lost, or melted in the heat of last weekend here in central and
south-western Europe. Why are only three names in the debug output, those whose
symlinks were created?

It would be helpful to run 'make -p' and 'make -d' from the corresponding
directories, with all symlinks each time removed, on those architectures (i386,
x86_64, PPC64) where no symlink is found missing.

Maybe it would also be helpful to attach the corresponding Makefiles.

I think a way to stop the build process in <architecture>/libgcc would be to
edit its Makefile and remove from target LIGCC_LINKS for example
md-unwind-def.h. But this Makefile has a rather recent date and no Makefile.in.
It's created by some configure script, possibly <top-level>/libgcc/configure…

Reply via email to