Closed loganharbour closed 2 years ago
Before we merge I want to turn my commit into a fixup, but more importantly I want to hear back on whether this fixed the problem in Freya.
I usually love the enormity of the "Undefined Behavior can produce nasal demons" joke, but frankly, between "nasal demons" and "your code breaks horribly but only 0.01% of the time", I'd have greatly preferred a trip to the rhinologist.
Job Coverage on b92f220 wanted to post the following:
Coverage did not change
This comment will be updated on new commits.
Hitting this error during libmesh build after the final linking of libdbg
, libopt
, and libdevel
. Do I need to bootstrap or manually configure TIMPI within moose/libmesh/contrib
? I was under the impression that configuration was occurring on the libmesh install script level.
make[4]: Leaving directory '/Users/user/projects/moose/libmesh/build/contrib/fparser'
make[3]: Leaving directory '/Users/user/projects/moose/libmesh/build/contrib/fparser'
make[2]: Leaving directory '/Users/user/projects/moose/libmesh/build/contrib/fparser'
Making all in nanoflann
make[2]: Entering directory '/Users/user/projects/moose/libmesh/build/contrib/nanoflann'
make[2]: Nothing to be done for 'all'.
make[2]: Leaving directory '/Users/user/projects/moose/libmesh/build/contrib/nanoflann'
Making all in timpi
make[2]: Entering directory '/Users/user/projects/moose/libmesh/build/contrib/timpi'
make[2]: *** No rule to make target 'all'. Stop.
make[2]: Leaving directory '/Users/user/projects/moose/libmesh/build/contrib/timpi'
make[1]: *** [Makefile:1034: all-recursive] Error 1
make[1]: Leaving directory '/Users/user/projects/moose/libmesh/build/contrib'
make: *** [Makefile:32810: all-recursive] Error 1
Do I need to bootstrap
If you switched out the libMesh TIMPI hash (v1.8.1_bootstrapped) for the final hash here? Yeah, exactly. When we actually merge this into libMesh we'll make a 1.8.2 release tag and branch with a bootstrapped tag for the libMesh submodule, but right here there's no bootstrap files in git.
I'm going to test this on https://moosedevelopers.slack.com/archives/C010M4X8ZK2/p1642138858010400
Getting all green on Freya with this change!
This also fixes the failed assertions on the NS input
Fantastic. I'm making this 1.8.2 and moving it downstream.
Thanks folks. Sorry for the misunderstanding of std::remove_if
Hey, don't blame yourself for this. I wrote the too-fragile class that broke, I reviewed the PR that broke it, and before you submitted the PR I specifically answered your question about whether std::remove_if
was safe to use here with an overconfident "yes".
@cticenhour here's the fix. It adds an extra branch within a loop (which is what we tried to avoid 😢)