This is the mail archive of the
libstdc++@gcc.gnu.org
mailing list for the libstdc++ project.
Re: PATCH: Install two PCH files instead of one
>Benjamin/Geoff, I think I also now know why Benjamin doesn't see any
>fails in libstdc++-v3 (even though he too has dwarf-2 on linux,
>right?). They are non-deterministic w.r.t. available memory...
>Given the machine Benjamin has compared to this one...
Hmmm. Odd.
I do see the g++ fails in the pch tests, but not the libstdc++ ones. I
thought you had some fixes for that?
>Well, that patch is not my call. Agreed, incremental fixing is good.
>What I have in my tree clearly fixes some outstanding PCH failures
>without adding new failures.
Hey, that sounds good to me.
>Benjamin, as the "owner" of the PCH enablement bits, OK?
Hmmm. I can't get this to work for me, although I like this idea a lot.
%g++ -Winvalid-pch -g -O0 -include $bld/H-x86-gcc/include/c++/3.4/stdc++.h 1.cc
%g++ -Winvalid-pch -g -O2 -include $bld/H-x86-gcc/include/c++/3.4/stdc++.h 1.cc
cc1plus: warning: /mnt/hd/bld/H-x86-gcc/include/c++/3.4/stdc++.h.gch/O0g: not used because `__NO_INLINE__' not defined
I noticed you weren't using -include $(path)/stdc++.h. I didn't think
this worked (although I would certainly prefer it if it did): I thought
absolute paths only were allowed. Am I confused or doing something wrong?
The other thing I just realized is that all the generated PCH files
should really be in the target include directory, not the toplevel one.
-benjamin