Rework all tests that interact with use after return to correctly handle the case where the mode has been explicitly set to Never or Always.
Details
Diff Detail
- Repository
- rG LLVM Github Monorepo
Event Timeline
llvm/lib/Transforms/Instrumentation/AddressSanitizer.cpp | ||
---|---|---|
260–277 | we are going to have two conflicting flags. static cl::opt<AsanDetectStackUseAfterReturnMode> ClUseAfterReturn("asan-use-after-return"... |
llvm/lib/Transforms/Instrumentation/AddressSanitizer.cpp | ||
---|---|---|
266–268 | I added this one and the one below ('1') to avoid updating tests. |
llvm/lib/Transforms/Instrumentation/AddressSanitizer.cpp | ||
---|---|---|
260–277 | 0,1 is fine | |
266–268 | then maybe '2' as well? | |
271 | it should be ASAN_OPTIONS=detect_stack_use_after_return=1 | |
3322–3325 | this comment applies to inside of the "then" branch only Please insert corresponsing spseudo code for else branch: | |
3344–3356 | we porbably need entire thing in the else branch This is future oportunity to improve: make compiler-rt never return NULL so we can remove CreateICmpEQ and one branch | |
3358 | do you want to uncomment and keep it? | |
3366 | FakeStack is LLVM object which represent result of the function call | |
llvm/test/Instrumentation/AddressSanitizer/stack-poisoning.ll | ||
21 | this statements checks that there is no this line in after previous check | |
23 | It would be nice to check for CreateICmpEQ | |
llvm/test/Instrumentation/AddressSanitizer/stack_dynamic_alloca.ll | ||
22 | same |
I added this one and the one below ('1') to avoid updating tests.
Should I just update the tests and remove these?