fogsong233 wrote: > Here you can find more information: #169272 > > The crux of the issue is that we recognize the out of line definition as a > statement. We need to extend the logic as I've mentioned in the previous pull > request, however, with the slight difference that we need to find out what is > the exact parse interface that clang calls when actually parsing this. Since > the other PR is generated by AI, I find it unlikely to be done by the author. > Let's try to fix it here.
And now my change is to use SuppressAccessChecks. We need to check whether it's a statement or a declaration syntactically, not semantically. Therefore, access checks should be suppressed during this tentative phase. Actually, I want to make isCXXDeclarationStatement fully tentative (reverting all side effects/annotations). This won't affect ParseFunctionDefinition, but since we are reverting everything, we need to adjust the logic in ParseTopLevelStmtDecl to handle the raw tokens (like ::) correctly when the tentative parse returns false. https://github.com/llvm/llvm-project/pull/178842 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
