You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
It is unclear why this happens, but looks like clone() method leaves some indirect memory leak. According to the provided information it seems like some sort of cycle in the ptrs (as there is no direct leak) an the offenders seems to be ll::ReturnOp and Scope.
This is unrelated to tower per-se, because the leak can be reproduced by vast-front if a clone operation is inserted at the end of conversion passes.
It is not obvious where this bug originates from, candidates are (in no particular order):
We are using clone in a way we cannot
There is something wrong in the definitions of some of our operations that leads to an incomplete erase
There is a bug in upstream
Steps to Reproduce
Run the test/repl/raise.c with address sanitizer.
The text was updated successfully, but these errors were encountered:
VAST version
master
LLVM version
18
Operating system
CI
Description
It is unclear why this happens, but looks like
clone()
method leaves some indirect memory leak. According to the provided information it seems like some sort of cycle in the ptrs (as there is no direct leak) an the offenders seems to bell::ReturnOp
andScope
.This is unrelated to
tower
per-se, because the leak can be reproduced byvast-front
if a clone operation is inserted at the end of conversion passes.It is not obvious where this bug originates from, candidates are (in no particular order):
clone
in a way we cannotSteps to Reproduce
Run the
test/repl/raise.c
with address sanitizer.The text was updated successfully, but these errors were encountered: