[PATCH] D104052: [clang] Fix CallExpr dependence bit may not respecting all its arguments.

hokein closed this revision. hokein marked an inline comment as done. hokein added a comment. committed in 314e456dfe85f8b5c53b85a7d815f7d463fe02ef . Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION

[PATCH] D104052: [clang] Fix CallExpr dependence bit may not respecting all its arguments.

sammccall accepted this revision. sammccall added inline comments. This revision is now accepted and ready to land. Comment at: clang/include/clang/AST/Expr.h:2991 + /// ! the dependence bits might be stale after calling this setter, it is + /// *caller*'s responsibility to

[PATCH] D104052: [clang] Fix CallExpr dependence bit may not respecting all its arguments.

hokein added a comment. sorry for the delay, I lost the track. > I think I can also live with adding explicit recomputeDependence() calls in > the right places, and slapping a warning on setArg() that it might be needed. I'm inclined with this option. As you mentioned, there are two major

[PATCH] D104052: [clang] Fix CallExpr dependence bit may not respecting all its arguments.

hokein updated this revision to Diff 355469. hokein added a comment. refine the solution: introduce a separate method, and call it when needed. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D104052/new/ https://reviews.llvm.org/D104052 Files:

[PATCH] D104052: [clang] Fix CallExpr dependence bit may not respecting all its arguments.

sammccall added a comment. Ping - curious about your thoughts here! Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D104052/new/ https://reviews.llvm.org/D104052 ___ cfe-commits mailing list

[PATCH] D104052: [clang] Fix CallExpr dependence bit may not respecting all its arguments.

sammccall added a comment. This seems correct but it also seems like it makes setting all the args of a CallExpr into a quadratic operation. e.g. ConvertArgumentForCall and BuildResolvedCallExpr in SemaExpr seem to do this (not just missing/changed args) and are probably on common code paths.

[PATCH] D104052: [clang] Fix CallExpr dependence bit may not respecting all its arguments.

hokein created this revision. hokein added a reviewer: sammccall. hokein requested review of this revision. Herald added a project: clang. Before this patch, the dependence of CallExpr was only computed in the constructor, the dependence bits might not reflect truth -- some arguments might be not