Closed JCGoran closed 1 week ago
Attention: Patch coverage is 50.00000%
with 5 lines
in your changes missing coverage. Please review.
Project coverage is 85.37%. Comparing base (
c8d0833
) to head (e740c87
).
Files | Patch % | Lines |
---|---|---|
src/codegen/codegen_neuron_cpp_visitor.cpp | 50.00% | 5 Missing :warning: |
:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.
Logfiles from GitLab pipeline #217405 (:white_check_mark:) have been uploaded here!
Status and direct links:
Logfiles from GitLab pipeline #217552 (:white_check_mark:) have been uploaded here!
Status and direct links:
Please check how widely _nt
is used in VERBATIM code. Then we can decide if switching to nt
is a good idea.
@1uc let's just say the answer is "non-zero". Can we change coreNEURON codegen instead then (i.e. use option 2 from the description)?
You might want to check neurodamus-models before changing the tests, BBP has written MOD files that consist mostly of VERBATIM blocks. I assume they work with CoreNEURON (and NEURON). Therefore, there's a good chance that they use both _nt
and nt
depending on which simulator is targeted.
Using both grep -RiIl 'verbatim' | xargs grep '\<_nt\>'
and directly grep -RiI '\<nt\>'
reveals no matches in neurodamus-models
; any other repos that could potentially be problematic? Otherwise, if there's too many things that could potentially break, I'll go back to the drawing board.
Logfiles from GitLab pipeline #218158 (:white_check_mark:) have been uploaded here!
Status and direct links:
at_time
, an internal NEURON function, requires an instance ofNrnThread*
, which is implicit in the mod file, but was previously called with a non-existing argument name (parent function hadNrnThread* _nt
, while the call toat_time
had passednt
).Note that I am slightly annoyed at the possible fixes here: in NEURON codegen, all of the arguments passed to functions apparently start with an underscore, while for coreNEURON codegen all of them start without an underscore. This suggests a couple of possible fixes:
nt
instead of_nt
(this PR)_nt
instead ofnt
(requires a bit more work since the unit tests check the codegen directly)ImplicitArgumentVisitor
(or make another one, or add a function in the codegen) which correctly sets the right variable (either with or without an underscore) depending on the the code backend used for the codegen