[PATCH] D69428: [GlobalOpt] Remove valgrind specific hacks (revert r160529)

[Jon Chesterfield via Phabricator via llvm-commits]

JonChesterfield added a comment.

In D69428#2689270 <https://reviews.llvm.org/D69428#2689270>, @jyknight wrote:

> MaskRay's example does not show an improvement in LSAN behavior, but rather a new false positive. Memory which is still accessible at process termination is not a leak, so the source -- as written -- had no leak

This seems an unusual definition of memory leak.

Is it predicated on the assumption of a hosted environment which will perform the memory free on behalf of the application at termination, where said host+application in some sense can never leak memory?


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D69428/new/

https://reviews.llvm.org/D69428