Conversation
This works around an issue with oneapi, where `-check uninit` generally is not useful because it uses the Memory Sanitizer For intel classic, `-check all ` includes `-check uninit` but for oneapi, it does not Also - turn off -`fpe0` for now Related to ACCESS-NRI/ACCESS-OM3#94
|
Ill cherry pick into other components if this looks ok |
anton-seaice
force-pushed
the
build_flags
branch
from
49e4171 to
fc5c461
Compare
dougiesquire
left a comment
dougiesquire
left a comment
There was a problem hiding this comment.
I wonder if we should be making use of RelWithDebInfo for this?
|
What's the usual build type for the repo? And is that build-type common/mostly-common across all the other repos? |
|
The usual build type is Release for this repo and the other model-component repos. My thinking is - we always want these settings - there's no context where we need a Release build of minimum size and without debugging information. And so distinguishing between Release and RelWithDebInfo doesn't help, we would just end up moving to RelWithDebInfo being the new default |
dougiesquire
left a comment
dougiesquire
left a comment
There was a problem hiding this comment.
I guess I was thinking about what would be standard CMake practice. But everything you've said is true and I don't feel strongly at all.
I'm happy to merge as is if @manodeep is
|
I also don't have strong feelings. The software transformation team is meeting up next week (in-person in Canberra) to discuss requirements/principles for cmake setup - likely that default build types and always-set flags (esp. this |
3 participants
This sets -grecord-gcc-switches per ACCESS-NRI/ACCESS-OM3#199
and
-check all per ACCESS-NRI/ACCESS-OM3#94