[cfe-dev] RFC: A proposal to move toward using C++11 features in LLVM & Clang / bounding support for old host compilers

[Ben Pope]

On 12/11/13 03:11, dag at cray.com wrote:
> Eric Christopher <echristo at gmail.com> writes:
> 
>>     What is so hard about waiting an extra month to give people a
>>     chance to test the new toolchain?
>>
>> You could start now with compiling under C++11 mode. The autoconf and
>> (I think) cmake ability is there now. That should get you your extra
>> month - especially if you count when we started this discussion.
> 
> This is very frustrating.  I'm not sure why my message is being lost.
> We need to know the specific version of gcc to test (in our case, others
> are concerned about other toolchains).
> 
> In all likelihood everything's going to work perfectly.  But in the past
> we have run into buggy compilers, nonconforming code in our sources,
> etc. and it's a ton of work to track down the problem and fix it.
> That's why the window is important.
> 
> We can't start testing until we know what to test.

Then decide what to test.

I don't know which version you are currently using, but it seems
plausible that the pain is going to be fairly similar regardless of
which different version you jump to.

So lets lay down a concrete proposal; bump up to gcc-4.8.2.  The number
of known serious regressions (according to the bugzilla link on the
front page of gcc website) is basically the same as 4.7.3 (146 vs 142
respectively, as of now).  4.6 branch is no longer maintained so I can't
get a regression count, but it's not likely to go down any time soon.
(Wayback results from April 1st of the bugzilla search says 146 serious
regressions, that was 12 days before the release of 4.6.4).

Ben