astrelni marked 5 inline comments as done.
astrelni added a comment.
Sorry for the long delay.
I've reworked the template instantiation stuff a little bit yet again. Going to
still come back and comment with results of profiling but I think now this
shouldn't be much slower than if the template instantiations were filtered out
with the matchers.
================
Comment at: clang-tidy/abseil/UpgradeDurationConversionsCheck.cpp:153
+ // required so we provide only a warning.
+ std::sort(MatchedTemplateLocations.begin(), MatchedTemplateLocations.end());
+ for (SourceLocation Loc : MatchedInstantiationLocations) {
----------------
alexfh wrote:
> MatchedTemplateLocations can be a std::set<SourceLocation> or a
> std::unordered_set<unsigned> filled with SourceLocation::getRawEncoding().
> Same for MatchedInstantiationLocations, I suppose.
Ah thank you, I didn't look hard enough after not seeing a std::hash
specialization for SourceLocation. In that case we don't need the
onEndOfTranslationUnit or MatchedInstantiationLocations so those are now gone.
I added a test case for when a template is used in source before the body since
I wasn't 100% sure that the template would be traversed before the
instantiation.
https://reviews.llvm.org/D53830
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
