[Mesa-dev] [PATCH v3 1/2] configure+mesa/st: unify check for -std=c++11 support and enable accordingly

[Emil Velikov]

On 2 October 2017 at 15:04, Gert Wollny <gw.fossdev at gmail.com> wrote:
> Am Montag, den 02.10.2017, 13:37 +0100 schrieb Emil Velikov:
>> Hi Gert,
>>
>> On 19 September 2017 at 12:35, Gert Wollny <gw.fossdev at gmail.com>
>> wrote:
>> > Unify the CXX feature tests for C++11 support that is required for
>> > SWR, clover, and mesa/st/tests.
>> >
>>
>> Surely we'll have capable compiler in 90+% of the time, still asking
>> for C++11 for single test is an overkill.
>
> Hence my attempt to unify the test for the three cases where c++11 is
> already mandatory.
>
SWR and Clover have been around for ages and they absolutely require C++11.
While rebuilding most of Mesa + its tests (as illustrated in 2/2) only
for a single test is very wasteful.

Let's not do that, pretty please?

> [...]
>
>>
>> > ---
>> >  .travis.yml                                   |   1 +
>>
>> Unrelated/undocumented change?
>
> Sorry. slipped my attention.
>
>>
>> >  configure.ac                                  |  13 +-
>> >  m4/ax_cxx_compile_stdcxx.m4                   | 987
>> > ++++++++++++++++++++++++++
>>
>> Please don't add custom m4 unless absolutely needed.
>
> The advantage of this module is that it actually tests features of the
> required standard, and if the compiler enables them by default, no flag
> is needed. The result is cached and the test already support c++14 and
> c++17.
>
> I can also use the already established AX_CHECK_COMPILE_FLAG, and if
> you say so, I'll do it, but IMHO the required changes to configure.ac
> make the code look less clean.
>
I'm roughly aware of the advantages, although I'm also thinking of the
disadvantages.
Close to nobody is interested in reading or maintaining m4, that I
know of. Thus having a custom one in tree is a bad idea.


>>
>> > --- a/configure.ac
>> > +++ b/configure.ac
>> > @@ -301,8 +301,12 @@ AX_CHECK_COMPILE_FLAG([-
>> > Wall],                                 [CXXFLAGS="$CXXFL
>> >  AX_CHECK_COMPILE_FLAG([-fno-math-
>> > errno],                       [CXXFLAGS="$CXXFLAGS -fno-math-
>> > errno"])
>> >  AX_CHECK_COMPILE_FLAG([-fno-trapping-
>> > math],                    [CXXFLAGS="$CXXFLAGS -fno-trapping-
>> > math"])
>> >  AX_CHECK_COMPILE_FLAG([-
>> > fvisibility=hidden],                   [VISIBILITY_CXXFLAGS="-
>> > fvisibility=hidden"])
>> > +AX_CXX_COMPILE_STDCXX(11, noext, optional)
>> > +
>> >  AC_LANG_POP([C++])
>> >
>> > +AM_CONDITIONAL(HAVE_STD_CXX11, test "x$HAVE_CXX11" = "x1")
>> > +
>> >  # Flags to help ensure that certain portions of the code -- and
>> > only those
>> >  # portions -- can be built with MSVC:
>> >  # - src/util, src/gallium/auxiliary, rc/gallium/drivers/llvmpipe,
>> > and
>> > @@ -2234,9 +2238,7 @@ if test "x$enable_opencl" = xyes; then
>> >          AC_MSG_ERROR([cannot enable OpenCL without Gallium])
>> >      fi
>> >
>> > -    if test $GCC_VERSION_MAJOR -lt 4 -o $GCC_VERSION_MAJOR -eq 4
>> > -a $GCC_VERSION_MINOR -lt 7; then
>> > -        AC_MSG_ERROR([gcc >= 4.7 is required to build clover])
>> > -    fi
>> > +    AX_CXX_COMPILE_STDCXX(11, noext, mandatory)
>> >
>>
>> Why check again - you already know the result.
>
> The result is cached, so calling the macro again is of no run-time
> consequence. The alternative is to replace this one-liner with a three
> line if-then statement (same if I switch to AX_CHECK_COMPILE_FLAG).
>
Surely a simple if check is not that ugly?

-Emil