balazske added inline comments.
================ Comment at: clang-tools-extra/test/clang-tidy/checkers/bugprone-unhandled-exception-at-new.cpp:143 + f_est_noexcept_dependent_used<true>(); +} ---------------- balazske wrote: > aaron.ballman wrote: > > You have tests for placement new with `nothrow_t`, but I think other forms > > of placement new are also very interesting to test as those typically don't > > throw. > > > > Additionally, perhaps tests where the allocation functions have been > > replaced by the user (such as a class allocation function)? > I realized that the case of user-defined constructor or allocation function > allows to throw any kind of exception. So the check should be improved to > handle this case: Not rely on the syntax of new expression, instead check if > the called allocation function or the called constructor may throw, and if > yes, check what exceptions are possible. What is your opinion, is it a better > approach? > (And a function to collect all possible exceptions called from a function is > needed, `ExceptionAnalyzer` seems usable.) It looks like that the user is free to define custom `operator new` and any constructor called that may throw any exception. Even in the "nothrow" case it is possible to use a constructor that may throw? If we would analyze every possible throwable exception that may come out from a new-expression, the checker would end up almost in a general checker that checks for uncaught exceptions. At least it is easy to extend. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D97196/new/ https://reviews.llvm.org/D97196 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
