Re: [PULL 20/33] configure: handle host compiler in probe_target_compiler

[Paolo Bonzini <pbonzini@redhat.com>]

On 6/21/22 20:30, Richard Henderson wrote:
> On 6/21/22 09:58, Paolo Bonzini wrote:
>> On 6/20/22 18:41, Matheus Kowalczuk Ferst wrote:
>>> On 17/06/2022 07:12, Paolo Bonzini wrote:
>>>> Hi Matheus,
>>>>
>>>> could you please test the tests-tcg-next branch at
>>>> https://gitlab.com/bonzini/qemu?
>>> At be6090bcac10, it works if no BE toolchain is present. Otherwise, the
>>> script probes powerpc64-linux-gnu-gcc instead of the native tools for
>>> ppc64le-linux-user, and then do_compiler fails because the
>>> $target_cflags contains -mlittle-endian.
>>>
>>
>> So the problem in that case is that powerpc64-linux-gnu-gcc is _not_
>> biarch and thus does not support -mlittle-endian?  (I thought PPC
>> compilers were all 32/64-bit and bi-endian).
> 
> Nit: this is ppc64le-linux-gnu-gcc, built for gcc135, a power9 ppc64le 
> host.  It *does* support -mbig-endian, but not -m32.
> 
> At least gcc11 as release was not biarch, with no special configure 
> arguments.  I can try upgrading to the current gcc12 release to see if 
> that changed...

Ok, yesterday I had tested with RHEL and there the ppc64le compiler is 
bi-arch and bi-endian, but multilibs are disabled.  So it cannot build 
32-bit hosted binaries like those for tests/tcg, only freestanding 
binaries (vof).

On the other hand the powerpc64-linux-gnu-gcc binary from the cross-gcc 
package is theoretically multilib-friendly, but it cannot find the CRT 
files on the host because they are not in the .../le multilib subdirectory.

The simplest way out is to just test both native and cross.  There is 
already code to check for a working compiler in the tests/tcg stanza, 
and it can be moved to probe_target_compiler.

Paolo