895915 - Cross compile builds' c++ host compiler does not build in C++11 mode

Status: RESOLVED FIXED

(Firefox Build System :: General, defect)

Description

[(no longer active)]

See bug 895322 comment 26.

Comment 1

[(no longer active)]

Trevor, thanks for working on bug 896463!

With that fix in production, what do we need to do here?  Should we just modify configure.in to look for gcc and g++ before cc and c++?

Comment 2

[Trevor Saunders (:tbsaunde)]

(In reply to :Ehsan Akhgari (needinfo? me!) from comment #1)
> Trevor, thanks for working on bug 896463!
> 
> With that fix in production, what do we need to do here?  Should we just
> modify configure.in to look for gcc and g++ before cc and c++?

there are two things I'm aware of that we need to do
1. change the android mozconfigs to set HOST_CXX="/tools/gcc-blah/whatever/it/is"
2. change configure to require  the host compiler support c++11 and have it had std= to HOST_CXXFLAGS, my impression from talking to glandium this morning is he's working on this part and I'm happy to leave it to him :)

Comment 3

[(no longer active)]

Cool, thanks a lot!

Comment 4

[Mike Hommey [:glandium]]

Attachment: Enforce host gcc/clang support for C++11

I think Ted is not working this week.

This happens to work despite not changing the host compiler because, guess what, since we're so bad at setting the right flags for the host compiler, we omit -fno-rtti, which is what breaks C++11 support in gcc 4.4 with libstdc++. I don't have the courage to actually fix the entire compile flags set for the host compiler with autoconf, I'll take on the occasion to make some progress on my python-based configure, but that'll be a separate bug.

Comment 5

[Mike Hommey [:glandium]]

Comment on attachment 779711 [details] [diff] [review]
Enforce host gcc/clang support for C++11

For whichever comes first :)

Comment 6

[Gregory Szorc [:gps]]

Comment on attachment 779711 [details] [diff] [review]
Enforce host gcc/clang support for C++11

Review of attachment 779711 [details] [diff] [review]:
-----------------------------------------------------------------

AFAICT you simply hoisted code from configure.in into toolchain.m4. If there were actual changes, I didn't see them. If there were changes, could you point them out our upload a patch-hoist patch making them clearer?

Comment 7

[Mike Hommey [:glandium]]

Comment on attachment 779711 [details] [diff] [review]
Enforce host gcc/clang support for C++11

There is additional code in MOZ_CXX11 (the host c++11 part)

Comment 8

[Gregory Szorc [:gps]]

Comment on attachment 779711 [details] [diff] [review]
Enforce host gcc/clang support for C++11

Review of attachment 779711 [details] [diff] [review]:
-----------------------------------------------------------------

I see the new code now! And I just read my previous review comment. WTF. I deserve to be slapped for writing such an incoherent mess. Sorry about that.

::: build/autoconf/toolchain.m4
@@ +185,5 @@
> +        HOST_CXXFLAGS="$HOST_CXXFLAGS -std=gnu++0x"
> +
> +	_SAVE_CXXFLAGS="$CXXFLAGS"
> +	_SAVE_CPPFLAGS="$CPPFLAGS"
> +	_SAVE_CXX="$CXX"

There are some tabs here. Splinter is failing me.

@@ +189,5 @@
> +	_SAVE_CXX="$CXX"
> +        CXXFLAGS="$HOST_CXXFLAGS"
> +        CPPFLAGS="$HOST_CPPFLAGS"
> +        CXX="$HOST_CXX"
> +        AC_CACHE_CHECK(for host gcc c++0x headers bug without rtti,

90% of this seems like a copy paste from above. While we could possibly factor out common bits, it's probably not worth it. Might be worth a short comment above explaining that any new additions should be mirrored here.

Comment 9

[Mike Hommey [:glandium]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/f92a6dcf7361
https://hg.mozilla.org/integration/mozilla-inbound/rev/37b47ffd5a40

Comment 10

[Ryan VanderMeulen (PTO, back 6-April)]

https://hg.mozilla.org/mozilla-central/rev/f92a6dcf7361
https://hg.mozilla.org/mozilla-central/rev/37b47ffd5a40