Re: [PATCH, master+11][gdb/build] Add CXX_DIALECT to CXX

[Simon Marchi via Gdb-patches <gdb-patches@sourceware.org>]

On 2021-09-08 8:15 a.m., Tom de Vries wrote:
> Hi,
> 
> The problem reported in PR28318 is that when using a gcc version that (while
> supporting c++11) does not support c++11 by default, the configure test for
> std::thread support will fail.
> 
> If gdb would use the default AX_CXX_COMPILE_STDCXX from autoconf, then we'd
> have:
> ...
> CXX="g++ -std=gnu++11"
> ...
> and the test for std::thread support would succeed.
> 
> Instead, gdb uses a custom AX_CXX_COMPILE_STDCXX (see commit 0bcda685399)
> which does:
> ...
> CXX=g++
> CXX_DIALECT=-std=gnu++11
> ...
> 
> We could add $CXX_DIALECT to the test for std::thread support, but that would
> have to be repeated for each added c++ support test.
> 
> Instead, fix this by doing:
> ...
> CXX="g++ -std=gnu++11"
> CXX_DIALECT=-std=gnu++11
> ...
> 
> The code added in gdb/ax_cxx_compile_stdcxx.m4 is copied from the default
> AX_CXX_COMPILE_STDCXX from autoconf.

From autoconf-archive?

At first glance, it looks like CXX_DIALECT would not be needed anymore,
since it's only used to be appended to CXX in Makefiles.  So now it
would end up twice in the command-line, which is not harmful but not
useful.

However, if we look at the original explanation of why we have this
local modification:

  We need to tweak AX_CXX_COMPILE_STDCXX a bit though.  Pristine
  upstream AX_CXX_COMPILE_STDCXX appends -std=gnu++11 to CXX directly.
  That doesn't work for us, because the top level Makefile passes CXX
  down to subdirs, and that overrides whatever gdb/Makefile may set CXX
  to.  The result would be that a make invocation from the build/gdb/
  directory would use "g++ -std=gnu++11" as expected, while a make
  invocation at the top level would not.

  So instead of having AX_CXX_COMPILE_STDCXX set CXX directly, tweak it
  to AC_SUBST a separate variable -- CXX_DIALECT -- and use '$(CXX)
  (CXX_DIALECT)' to compile/link.

  https://gitlab.com/gnutools/binutils-gdb/-/commit/0bcda68539948828795564b35a497dc69c27f768

So it sounds like we need both (and your patch is correct):

 - the switch in CXX for the std::thread tests (and other tests)
 - the switch in CXX_DIALECT so it can be appended in Makefiles, to
   counteract the fact that the top-level Makefile overrides CXX

This should probably be explained in the comment, because that situation
is strange, and people might wonder why we end up with two -std=...
switches, and might try to "fix" it.

Simon