Fix an assertion when building the clad benchmarks with clang18 on osx.

[vgvassilev]

Unfortunately, it'd be difficult to come up with a test...

[github-actions]

clang-tidy review says "All clean, LGTM! 👍"

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 93.72%. Comparing base (b38d8cf) to head (6bfae40).
Report is 8 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #930      +/-   ##
==========================================
- Coverage   94.08%   93.72%   -0.36%     
==========================================
  Files          53       54       +1     
  Lines        7762     7782      +20     
==========================================
- Hits         7303     7294       -9     
- Misses        459      488      +29     

Files	Coverage Δ
lib/Differentiator/VisitorBase.cpp	96.65% <100.00%> (ø)

... and 6 files with indirect coverage changes

Files	Coverage Δ
lib/Differentiator/VisitorBase.cpp	96.65% <100.00%> (ø)

... and 6 files with indirect coverage changes

[vaithak]

Looks good.
But, given that we will improve the error or diagnostic messages in the future, maybe we can use E->getBeginLoc and E->getEndLoc instead of creating fake ones.

[vgvassilev]

Looks good. But, given that we will improve the error or diagnostic messages in the future, maybe we can use E->getBeginLoc and E->getEndLoc instead of creating fake ones.

Until now we explicitly tried to use the fakeLoc approach to consistently annotate where we need to handle source locations in a non-crashing way. We could but that would point to the original function. Is that what you want here?

[vaithak]

Until now we explicitly tried to use the fakeLoc approach to consistently annotate where we need to handle source locations in a non-crashing way. We could but that would point to the original function. Is that what you want here?

Although it would point to the original function, it will be really helpful for creating minimal reproducers from big functions, as these location informations in the primal function will point to the line or location of the troublesome expression.

[vgvassilev]

Until now we explicitly tried to use the fakeLoc approach to consistently annotate where we need to handle source locations in a non-crashing way. We could but that would point to the original function. Is that what you want here?

Although it would point to the original function, it will be really helpful for creating minimal reproducers from big functions, as these location informations in the primal function will point to the line or location of the troublesome expression.

Ok. Let me change that. However, this comment is questioning our currently strategy here, right? If you agree we should open an issue where we suggest replacing all of the fake locations with ones pointing to the original function.

[vaithak]

Ok. Let me change that. However, this comment is questioning our currently strategy here, right? If you agree we should open an issue where we suggest replacing all of the fake locations with ones pointing to the original function.

I am in full support for the new strategy of pointing to original function locations, so that it provides ease of debugging.

[github-actions]

clang-tidy review says "All clean, LGTM! 👍"

[vgvassilev]

I do not think the codecov complaint is relevant for this PR. Let's move forward.