>> On 8 Jun 2026, at 15:23, Pietro Monteiro <[email protected]> wrote:
>>
>> On Mon, Jun 8, 2026, at 9:59 AM, Iain Sandoe wrote:
>>>> On 8 Jun 2026, at 14:53, Jose E. Marchesi <[email protected]> wrote:
>>>>
>>>>>>
>>>>>> On 8 Jun 2026, at 14:09, Pietro Monteiro <[email protected]>
>>>>>> wrote:
>>>>>>
>>>>>> Mach-O objects prefix section names with double underscore instead of
>>>>>> period. Use correct exports section name on systems that use Mach-O.
>>>>>>
>>>>>> gcc/algol68/ChangeLog:
>>>>>>
>>>>>> * a68.h (A68_EXPORT_SECTION_NAME): Prefix name with double
>>>>>> underscores for Mach-O objects.
>>>>>
>>>>> I am a bit puzzled by this - since I had memory that we (Jose and I) had
>>>>> already
>>>>> fixed things so that Darwin’s segment and section names were correct.
>>>>> (indeed, AFAIK a68 exports in object files and convenience libraries are
>>>>> already
>>>>> working)?
>>>>
>>>> I am also a bit confused. From our past testing, it looked like in
>>>> Darwin it was ok to create a section .a68_exports and then read it in,
>>>> conventions notwithstanding.
>>>
>>> That is my expectation - that, at the binary level, all that matters is
>>> that the
>>> segment and section names are <= 16 chars in length (there is no terminating
>>> nul in the file).
>>
>> Without the patch `otool` shows the libga68 .dylib having a __GNU_A68
>> segment with a __a68_exports section in it. Maybe the system `as`
>> (clang) "fixes" the name ?
>
> I don’t think so.
>
> e.g. posix.s ==> posix.o
>
> .build_version macos, 15, 0
> .section __GNU_A68,__a68_exports
>
> It looks like we are emitting this from the compiler already .. which kinda
> stacks
> up with my recollection that we’d already been down this road - I do apologise
> that my recollection is fuzzy ..
So given we do:
static void
a68_switch_to_export_section (void)
{
static section *exports_sec;
if (exports_sec == NULL)
{
gcc_assert (targetm_common.have_named_sections);
#ifdef OBJECT_FORMAT_MACHO
exports_sec
= get_section (A68_EXPORT_SEGMENT_NAME "," A68_EXPORT_SECTION_NAME,
SECTION_DEBUG, NULL);
#else
exports_sec = get_section (A68_EXPORT_SECTION_NAME,
TARGET_AIX_OS ? SECTION_EXCLUDE :
SECTION_DEBUG,
NULL);
#endif
}
switch_to_section (exports_sec);
}
It must be either get_section that is turning __GNU_A68,.a68_exports
into __GNU_A68,__a68_exports, or there is something down the pipeline
doing that subst in macos targets..
>
> Iain
