[PATCH] D35443: Enable 64-bit Darwin LeakSanitizer by default on AddressSanitizer builds

[Francis Ricci via Phabricator via llvm-commits]

fjricci added a comment.

Hi @kubamracek - turns out that the llvm-tblgen issues were caused by DCE leading to a broken weak hook, which I fixed here: https://reviews.llvm.org/D37636.

Bootstrapped check-clang reports only 2 test failures with this patch. One is a known leak in this patch, which I've tried fixing, but haven't had much luck on since I'm not familiar with the codebase: https://reviews.llvm.org/D34784. Presumably that leak doesn't show up on buildbots currently because the test is disabled on Linux. The other failing test also doesn't run on linux, and is a forced crash test (crash-vfs-ivfsoverlay.m). Perhaps that one is failing because LSan changes exit behavior? LSan doesn't actually report a false positive leak there, it just omits some of the crash dump printout.

What's the preferred way to handle these failures? I think the current state demonstrates that LeakSanitizer itself works well. How important is swift support? I haven't had much time to look into that, since hasn't been a big priority for me.


https://reviews.llvm.org/D35443