830315 - MOZ_FINAL and MOZ_DELETE cause a warning in clang in C++98 mode

Status: RESOLVED WORKSFORME

(Core :: MFBT, defect)

Description

[Mike Hommey [:glandium]]

warning: deleted function definitions are a C++11 extension [-Wc++11-extensions]
warning: 'final' keyword is a C++11 extension [-Wc++11-extensions]

Apparently, __has_extension(cxx_deleted_functions) and __has_extension(cxx_override_control) are still true when __cplusplus < 201103L.

Comment 1

[Mike Hommey [:glandium]]

Attachment: Don't enable MOZ_FINAL and MOZ_DELETE on clang unless compiling as C++11

Comment 2

[Jeff Walden [:Waldo]]

Comment on attachment 701774 [details] [diff] [review]
Don't enable MOZ_FINAL and MOZ_DELETE on clang unless compiling as C++11

Review of attachment 701774 [details] [diff] [review]:
-----------------------------------------------------------------

Odd.  We compile SpiderMonkey as C++98, and use override all the time there, so I'm not sure why I've never seen these warnings there.  Oh well, these warnings obviously manifest on a test program, so this is the right thing to do for now.

I'd hoped __extension__ override or __extension__(override) would work to silence the warning and give override semantics, but no dice for now.  Maybe we'll get it in some future version of clang; I filed http://llvm.org/bugs/show_bug.cgi?id=14953 on that.

Comment 3

[Mike Hommey [:glandium]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/1c7bdeefa8ee

Comment 4

[Ed Morley [:emorley]]

https://hg.mozilla.org/mozilla-central/rev/1c7bdeefa8ee

Comment 5

[Ryan VanderMeulen [:RyanVM]]

This was backed out in bug 831633.
https://hg.mozilla.org/mozilla-central/rev/88fc571a9e9e

Comment 6

[Seth Fowler [:seth] [:s2h]]

It'd be _really_ nice to get this resolved. This spews out a huge number of warnings, and if you build at a terminal, the combination of clang's colorized output and the interleaving caused by parallel builds makes my terminal program repeatedly bring up print dialogs. Sometimes it also gets the terminal in an unusable state.

Comment 7

[Mike Hommey [:glandium]]

You shouldn't actually be seeing these particular warnings, as mentioned in bug 831633. They should only appear for HOST_C*SRCS files, and we don't build that many.

Comment 8

[Jeff Walden [:Waldo]]

I assume you're going to add the relevant warning flags to HOST_*FLAGS, right?  I can't think of a reason why we should have different flags for host versus target here.

Comment 9

[Seth Fowler [:seth] [:s2h]]

I agree that I shouldn't be seeing them. =) I see a lot of them, though, on a clean build (OS X 10.8). IIRC the JS engine is the primary source of them.

Comment 10

[Seth Fowler [:seth] [:s2h]]

I don't see this happening anymore.