Auditing printf-likes in bug 1060419 revealed that DMD could use MOZ_FORMAT_PRINTF in a few spots, namely the various implementations of StatusMsg.

IIUC to run this in try I have to also add a patch to update some moz config files.

Comment on attachment 8862053 [details] Bug 1334286 - use MOZ_FORMAT_PRINTF in DMD; https://reviewboard.mozilla.org/r/134006/#review137014 ::: memory/replace/dmd/DMD.h:53 (Diff revision 1) > > virtual void Analyze(UniquePtr<JSONWriteFunc>); > > virtual void SizeOf(Sizes*); > > - virtual void StatusMsg(const char*, va_list); > + virtual void StatusMsg(const char*, va_list) MOZ_FORMAT_PRINTF(2, 0); Is 0 the second arg here because it takes a va_list instead of ...? The docs in mfbt/Attributes.h don't mention that case.

(In reply to Tom Tromey :tromey from comment #2) > IIUC to run this in try I have to also add a patch to update some moz config > files. You're in luck. DMD is enabled for debug builds on TreeHerder -- there is even an xpcshelltest -- so you won't have to do that.

(In reply to Nicholas Nethercote [:njn] from comment #3) > > + virtual void StatusMsg(const char*, va_list) MOZ_FORMAT_PRINTF(2, 0); > > Is 0 the second arg here because it takes a va_list instead of ...? Yes. See also bug 1334279. This is part of GCC's definition of the attribute.

Pushed by ttromey@mozilla.com: https://hg.mozilla.org/integration/autoland/rev/9c090ad2183e use MOZ_FORMAT_PRINTF in DMD; r=njn