https://gcc.gnu.org/bugzilla/show_bug.cgi?id=128414
Tobias Burnus <burnus at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Resolution|--- |INVALID
Status|WAITING |RESOLVED
--- Comment #2 from Tobias Burnus <burnus at gcc dot gnu.org> ---
Close as INVALID ...
(In reply to Thomas Schwinge from comment #1)
> Based on similar such things that you've reported in the past: please check
> that your GCC/nvptx' 'mgomp' multilib exists.
I bootstrapped now – and it works now. [I probably should do this
more often, but a full bootstrap takes ~50 min per worktree such
that I don't want to do this all the time. I need a faster laptop ;-)]
> In case my suspicion from above is right, I wonder what you're doing to get
> your GCC/nvptx into that broken state, every once in a while?
Good question. I just do incremental builds with clean bootstraps once in a
while and just
make -j24 && make install
in the gcn, nvptx, and host build directory.
I think 'rm -rf */libgomp/' [+ libgomp multilibs] and then rebuilding helped in
some cases with libgomp issues – but here it didn't.
[I don't know whether it has ever helped with 'mgomp' issues or not. Nor do I
understand why there are nvptx issues; there do not seem to be such issues with
GCN.]
