ompi: rename assert variables to mpi_assert #8591
Conversation
There already exists an identifier with the name assert. It is not appropriate to use this name for variables in C. This commit changes these variable names to mpi_assert. This should address an issue where clang-format gives the wrong result when doing bit-wise-and operations on these variables. Signed-off-by: Nathan Hjelm <[email protected]>
hjelmn
force-pushed
the
for_the_love_of_all_things_DO_NOT_NAME_VARIABLES_ASSERT
branch
from
9387dab to
aaabe3d
Compare
|
bot:aws:retest one more time... |
There was a problem hiding this comment.
Fine with me since it works around an issue with clang-format.
For the record: using assert as a variable name is perfectly. The C assert is a function-like macro that is only expanded when called. Otherwise the pre-compiler will not touch it. See https://gcc.gnu.org/onlinedocs/cpp/Function-like-Macros.html
|
@devreal That makes sense. Wasn't sure how it is defined but the C11 standard clearly says |
|
Not sure what is going on with the build checker. Kicking it again bot:aws:retest |
|
@hjelmn There is a |
|
@awlauria That sounds like a good idea. Generally, shadowing leads to inadvertent errors. If gcc is not already warning we should turn that on. |
|
Though thinking about it |
|
Yeah unfortunately it won't pick up the assert() case, but still a good thing to scrub. |
|
Ugh. The build checker is busted. |
|
Looked the build checker and everything worked fine except for a random failure in the clang38 workflow. Not related to this change. |
There already exists an identifier with the name assert. It is not
appropriate to use this name for variables in C. This commit changes
these variable to mpi_assert. This should address an issue where
clang-format gives the wrong result when doing bit-wise operations on
these variables.
Signed-off-by: Nathan Hjelm [email protected]