Hi!

On 2023-06-03T21:32:57+0100, "Maciej W. Rozycki" <macro@orcam.me.uk> wrote:
>> Will you, Maciej, please test that this doesn't break your setting?
>
>  Umm, this was implemented for my Western Digital development environment,
> which I don't have access to anymore.  I'll see what I can do, but it may
> be neither easy nor quick.  It's been long ago and I don't have a setup
> with multilibs enabled anymore.  And neither I remember the thorough
> problem analysis I went through that has led me to the conclusions.

I see.  I've therefore myself now done a quick hack to replicate the
original requirement, to verify that given '--with-build-sysroot=',
build-tree testing must have '--sysroot=[...]' appear for every driver
invocation:

    --- gcc/gcc.cc
    +++ gcc/gcc.cc
    @@ -8190,6 +8190,10 @@ driver::main (int argc, char **argv)
       if (!maybe_print_and_exit ())
         return 0;

    +  if (!env.get ("SKIP_VERIFY_SYSROOT"))
    +    if (!(target_system_root && !strcmp (target_system_root, "/boot/..")))
    +      internal_error ("MISSING SYSROOT");
    +
       early_exit = prepare_infiles ();
       if (early_exit)
         return get_exit_code ();

With that, build GCC with '--with-build-sysroot=/boot/..' and with the
environment variable 'SKIP_VERIFY_SYSROOT' set (to ignore any build-time
issues, not relevant to this discussion here), and test without the
'SKIP_VERIFY_SYSROOT' environment variable set (meaning the checking is
active).  I observe that with current (un-altered) GCC sources,
'--sysroot=[...]' doesn't actually appear for all compiler test suites;
the following ones FAIL:

  - 'ada/acats/acats.sum'
  - 'testsuite/g++/g++.sum': 'g++.dg/plugin/plugin.exp' only
  - 'testsuite/gcc/gcc.sum': 'gcc.dg/plugin/plugin.exp' only
  - 'gm2/gm2.sum'
  - 'gnat/gnat.sum'
  - 'obj-c++/obj-c++.sum': 'obj-c++.dg/plugin/plugin.exp' only
  - 'objc/objc.sum': 'objc.dg/plugin/plugin.exp' only

Additionally, the following target library test suite also FAILs:

    'libitm/testsuite/libitm.sum'

(Resolving these is not my objective right now; I suppose these were not
relevant in Maciej's original scenario.)

Otherwise, I observe that my proposed re-work does still achieve the
desired outcome re '--sysroot=[...]', and Iain has long ago confirmed
that it does resolve <https://gcc.gnu.org/PR109951>
"libgomp, testsuite: non-native multilib c++ tests fail on Darwin", so
pushed to master branch commit fb5d27be272b71fb9026224535fc73f125ce3be7
"libgomp: Consider '--with-build-sysroot=[...]' for target libraries' build-tree testing (instead of build-time 'CC' etc.) [PR91884, PR109951]",
see attached.


Grüße
 Thomas


-----------------
Siemens Electronic Design Automation GmbH; Anschrift: Arnulfstraße 201, 80634 München; Gesellschaft mit beschränkter Haftung; Geschäftsführer: Thomas Heurung, Frank Thürauf; Sitz der Gesellschaft: München; Registergericht München, HRB 106955