Fix function definition regex bug

[gregersrygg]

The whitespace fix to this regex last week triggered another bug in
libmodem. They have a typedef that the regex parsed as a function
definition because it accepted too many character in the argument
list of a function. This fix updates the regex to only allow
characters you would expect in an argument list of a function def.

Characters expected in the argument list of a function def:

\w A-Z a-z 0-9 and _
\s All whitespace
,  Argument separator
*  Pointers
\. Variadic argument
\[ Array start
\] Array end

[NordicBuilder]

You can find the documentation preview for this PR at this link. It will be updated about 10 minutes after the documentation build succeeds.

Note: This comment is automatically posted by the Documentation Publishing GitHub Action.

[lemrey]

Thanks @gregersrygg, mocks are generated correctly in libmodem now.
I wonder if the CI on this PR runs any unit tests though, we should make sure it does.

[lemrey]

@nordic-krch , @tejlmand please have a look.

[tejlmand]

Could you please add a test for your new regex here with an example of what you're fixing:
https://github.com/nrfconnect/sdk-nrf/blob/main/tests/unity/example_test/src/foo/foo.h

feel free to add such tests in other files under https://github.com/nrfconnect/sdk-nrf/blob/main/tests/unity/example_test if more appropriate.

Please also add a test for the previous fix you made here: #6478

[gregersrygg]

@tejlmand Yes, I noticed the example_test while going through previous commits on some of the files I touched. I also felt the need for more testing on this functionality. However I didn't want to clutter the example_test with more tests, so I separated the wrap tests from the example_test.

[tejlmand]

However I didn't want to clutter the example_test with more tests, so I separated the wrap tests from the example_test.

Perfectly fine.

Just wanted to ensure to get some test on this, so thanks for improving testing.

[balaji-srin]

I think it is better to write a README or some documentation explaining what example_test can be used for.

[gregersrygg]

I also made a small experiment on using pycparser to parse the C code instead of using regexes. However it relies on the C preprocessor to run first (for #includes), so the cmake script would have to pass include paths to the python script, and it would also be difficult to output the modified C files with the #includes reverted back to how it was before the preprocessor. So it's not as easy as I had hoped.

[gregersrygg]

@balaji-nordic
I think it is better to write a README or some documentation explaining what example_test can be used for.

This PR only removes code from the example_test, so I don't understand why I would then have to write documentation for for it as part of this bugfix/cleanup. There's also end user documentation for the example_test on our documentation site.

[tejlmand]

lgtm.

Approved under the expectation that snippets.c/h will be renamed according to discussions.

[gregersrygg]

The compliance check that fail is because of intentional syntax in tests/unity/wrap_test/src/test_code/test_code.h. It is formatted this way to mimic syntax that has broken the code parsing in the past:

• tests: unity: Cleanup header parsing regexes #1413
• scripts: unity: Fix header_prepare comments removal regex #2182
• scripts: unity: header_prepare.py: Improve regex #3432
• Fixes in unity framework #5069
• scripts: unity: func_name_list: Improve function name search #5814
• scripts: unity: Fix whitespace not handled correctly #6478

[tejlmand]

@carlescufi or @mbolivar-nordic this is good to merge.

Compliance failures are expected, as test code need to break formatting to test that the scripts works also on 3rd party code that are not following our coding guideline (and hence our test headers replicating that formatting break our guidelines)