[Bug ada/112958] New: [14 regression] s-exnllf.ads etc. don't compile on 32-bit FreeBSD/x86

[Bug ada/112958] New: [14 regression] s-exnllf.ads etc. don't compile on 32-bit FreeBSD/x86

[ro at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112958

            Bug ID: 112958
           Summary: [14 regression] s-exnllf.ads etc. don't compile on
                    32-bit FreeBSD/x86
           Product: gcc
           Version: 14.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: ada
          Assignee: unassigned at gcc dot gnu.org
          Reporter: ro at gcc dot gnu.org
                CC: andreast at gcc dot gnu.org, dkm at gcc dot gnu.org
  Target Milestone: ---
              Host: x86_64-unknown-freebsd14.0
            Target: x86_64-unknown-freebsd14.0
             Build: x86_64-unknown-freebsd14.0

On current trunk, at least libgnat sources don't compile on 32-bit FreeBSD/x86:

s-exnllf.ads:38:04: warning: in instantiation at s-dorepr.adb:82 [enabled by
def
ault]
s-exnllf.ads:38:04: warning: in instantiation at s-exponr.adb:54 [enabled by
def
ault]
s-exnllf.ads:38:04: warning: "Temp" overlays smaller object [enabled by
default]
s-exnllf.ads:38:04: warning: program execution may be erroneous [enabled by
defa
ult]
s-exnllf.ads:38:04: warning: size of "Temp" is 96 [enabled by default]
s-exnllf.ads:38:04: warning: size of "Rep64" is 64 [enabled by default]
make[9]: *** [../gcc-interface/Makefile:306: s-exnllf.o] Error 1

-imgllf.ads:49:04: warning: in instantiation at s-dorepr.adb:82 [enabled by def
ault]
s-imgllf.ads:49:04: warning: in instantiation at s-imager.adb:61 [enabled by
def
ault]
s-imgllf.ads:49:04: warning: "Temp" overlays smaller object [enabled by
default]
s-imgllf.ads:49:04: warning: program execution may be erroneous [enabled by
defa
ult]
s-imgllf.ads:49:04: warning: size of "Temp" is 96 [enabled by default]
s-imgllf.ads:49:04: warning: size of "Rep64" is 64 [enabled by default]
make[9]: *** [../gcc-interface/Makefile:309: s-imgllf.o] Error 1

I suspect this is due to to FreeBSD/x86's unusual long double format in the
32-bit case: i386/freebsd.h has

#define TARGET_96_ROUND_53_LONG_DOUBLE (!TARGET_64BIT)

[Bug ada/112958] [14 regression] s-exnllf.ads etc. don't compile on 32-bit FreeBSD/x86

[ro at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112958

Rainer Orth <ro at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |14.0

[Bug ada/112958] [14 regression] s-exnllf.ads etc. don't compile on 32-bit FreeBSD/x86

[ebotcazou at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112958

Eric Botcazou <ebotcazou at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|                            |2024-01-09
     Ever confirmed|0                           |1
                 CC|                            |ebotcazou at gcc dot gnu.org
             Status|UNCONFIRMED                 |WAITING

--- Comment #1 from Eric Botcazou <ebotcazou at gcc dot gnu.org> ---
The code is the same on the 13 branch though, does it compile there?

[Bug ada/112958] [14 regression] s-exnllf.ads etc. don't compile on 32-bit FreeBSD/x86

[ro at CeBiTec dot Uni-Bielefeld.DE]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112958

--- Comment #2 from ro at CeBiTec dot Uni-Bielefeld.DE <ro at CeBiTec dot Uni-Bielefeld.DE> ---
> --- Comment #1 from Eric Botcazou <ebotcazou at gcc dot gnu.org> ---
> The code is the same on the 13 branch though, does it compile there?

So far, I had only tried gcc 11.4.0 (where the code compiles) and trunk
(where it doesn't).  Now, I've tried both the gcc-12 and gcc-13 branches
and both fail like trunk.

[Bug ada/112958] [12/13/14 regression] s-exnllf.ads etc. don't compile on 32-bit FreeBSD/x86

[ebotcazou at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112958

Eric Botcazou <ebotcazou at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[14 regression]             |[12/13/14 regression]
                   |s-exnllf.ads etc. don't     |s-exnllf.ads etc. don't
                   |compile on 32-bit           |compile on 32-bit
                   |FreeBSD/x86                 |FreeBSD/x86
             Status|WAITING                     |NEW

--- Comment #3 from Eric Botcazou <ebotcazou at gcc dot gnu.org> ---
Thanks.

[Bug ada/112958] [12/13/14 regression] s-exnllf.ads etc. don't compile on 32-bit FreeBSD/x86

[ebotcazou at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112958

--- Comment #4 from Eric Botcazou <ebotcazou at gcc dot gnu.org> ---
Someone motivated enough should add a specific libgnat/s-dorepr__freebsd.adb
unit where Rep64 is an array of two Interfaces.Unsigned_64 and the AND
operation on line 88 is performed on the first component.

[Bug ada/112958] [12/13/14 regression] s-exnllf.ads etc. don't compile on 32-bit FreeBSD/x86

[ro at CeBiTec dot Uni-Bielefeld.DE]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112958

--- Comment #5 from ro at CeBiTec dot Uni-Bielefeld.DE <ro at CeBiTec dot Uni-Bielefeld.DE> ---
> --- Comment #4 from Eric Botcazou <ebotcazou at gcc dot gnu.org> ---
> Someone motivated enough should add a specific libgnat/s-dorepr__freebsd.adb
> unit where Rep64 is an array of two Interfaces.Unsigned_64 and the AND
> operation on line 88 is performed on the first component.

Thanks for the analysis.  I guess I'll leave that to the FreeBSD folks
then: this was just one of many issues I found during a trunk FreeBSD
bootstrap.

[Bug ada/112958] [12/13/14/15 regression] s-exnllf.ads etc. don't compile on 32-bit FreeBSD/x86

[gerald at pfeifer dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112958

--- Comment #6 from Gerald Pfeifer <gerald at pfeifer dot com> ---
FreeBSD i386 is on it's way out: FreeBSD 14 is the last series supporting
it; FreeBSD 15 is dropping support for it.

Can we disable libgnat (or GNAT) support during configure time?

[Bug ada/112958] [12/13/14/15 regression] s-exnllf.ads etc. don't compile on 32-bit FreeBSD/x86

[ro at CeBiTec dot Uni-Bielefeld.DE]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112958

--- Comment #7 from ro at CeBiTec dot Uni-Bielefeld.DE <ro at CeBiTec dot Uni-Bielefeld.DE> ---
> --- Comment #6 from Gerald Pfeifer <gerald at pfeifer dot com> ---
> FreeBSD i386 is on it's way out: FreeBSD 14 is the last series supporting
> it; FreeBSD 15 is dropping support for it.

Ah, I wasn't aware of that (and was particularly interested in i386 for
comparison with Solaris/i386 asan).

> Can we disable libgnat (or GNAT) support during configure time?

Sure: --disable-libasan should work for every toplevel library, although
there's little point in building GNAT without the runtime libs.  Given
that there's no --disable-languages (I've filed a PR for that ages ago,
I believe), you have to use

        --enable-languages=<list of every language you want to build
                            execept for ada>

instead of

        --enable-languages=all --disable-languages=ada

[Bug ada/112958] [12/13/14/15 regression] s-exnllf.ads etc. don't compile on 32-bit FreeBSD/x86

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112958

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|14.0                        |14.2

--- Comment #8 from Richard Biener <rguenth at gcc dot gnu.org> ---
GCC 14.1 is being released, retargeting bugs to GCC 14.2.