Hi Ian,
> On Fri, Feb 15, 2019 at 4:03 AM Rainer Orth <[email protected]>
> wrote:
>>
>> Andreas Schwab <[email protected]> writes:
>>
>> > This breaks non-split-stack builds.
>> >
>> > ../../../libgo/runtime/stack.c: In function 'doscanstack1':
>> > ../../../libgo/runtime/stack.c:113:18: error: passing argument 1 of
>> > 'scanstackblock' makes integer from pointer without a cast
>> > [-Werror=int-conversion]
>> > 113 | scanstackblock(bottom, (uintptr)(top - bottom), gcw);
>> > | ^~~~~~
>> > | |
>> > | byte * {aka unsigned char *}
>>
>> I see the same on Solaris. Even with that fixed by appropriate casts to
>> uintptr (plus a few more times), Solaris bootstrap is still broken by
>> that patch:
>>
>> /vol/gcc/src/hg/trunk/local/libgo/runtime/go-varargs.c: In function
>> '__go_syscall6':
>> /vol/gcc/src/hg/trunk/local/libgo/runtime/go-varargs.c:101:10: error:
>> implicit declaration of function 'syscall'
>> [-Werror=implicit-function-declaration]
>> 101 | return syscall (flag, a1, a2, a3, a4, a5, a6);
>> | ^~~~~~~
>>
>> This needs to include <sys/syscall.h> for the syscall declaration, apart
>> from the fundamental problem that syscall isn't a stable interface on
>> Solaris.
>
> I committed this patch which should fix the Solaris build.
it did, and certainly in a cleaner way than just unconditionally
including <sys/syscall.h> ;-) Thanks.
> The code was already calling syscall, it was just doing it in a way
> that the types didn't necessarily match the C declaration. This is
> the implementation of Go's syscall.Syscall function, so there isn't
> really anything else we can do.
I feared as much. Some time ago when debugging another issue I saw
libgo using syscall() directly, certainly unexpected in that particular
case.
Rainer
--
-----------------------------------------------------------------------------
Rainer Orth, Center for Biotechnology, Bielefeld University
