On Thu, Nov 28, 2019 at 05:45:06PM +0000, Ramsay Jones wrote:
>
> Signed-off-by: Ramsay Jones <ramsay@xxxxxxxxxxxxxxxxxxxx>
> ---
>
> Hi Luc,
Hi,
> This is marked RFC because it only improves the situation on 64-bit cygwin.
> Without access to a (up-to-date) 32-bit cygwin, I can't experiment to find
> a means to determine what platform I am on. I don't recall what the output
> of 'uname' is on 32-bit cygwin, but I have a hunch that you can't tell which
> is which from it's output. On 64-bit cygwin:
>
> $ uname -a
> CYGWIN_NT-10.0 satellite 3.0.7(0.338/5/3) 2019-04-30 18:08 x86_64 Cygwin
> $ uname -s
> CYGWIN_NT-10.0
> $ uname -o
> Cygwin
> $
>
> [ie. I don't think 'uname -o' returns Cygwin32 or similar. :( ]
Indeed, I'm guess it doesn't.
> So, I don't know.
I see several possibilities:
1) just this patch, wich is OK for 64-bit platform/compiler
where 32-bit needs to be forced with -m32
2) simply not define __CYGWIN32__ at all based on the
conviction that it's only __CYGWIN__ that should be tested
3a) in cgcc add 'm32=1' in the i386 part if $m64 is not set
3b) in cgcc add 'm64=1' in the x86_64 part if $m32 is not set
and change this patch to test $m64 instead of testing $m32
4a) in sparse itself, add something like:
if (arch_mach == MACH_X86_64 && arch_os == OS_CYGWIN)
add_pre_buffer("#undef __CYGWIN32__");
or:
if (arch_m64 != LP32 && arch_os == OS_CYGWIN)
add_pre_buffer("#undef __CYGWIN32__");
4b) do not define __CYGWIN32__ in cgcc and add something like:
if (arch_mach == MACH_i386 && arch_os == OS_CYGWIN)
add_pre_buffer("#define __CYG_WIN32__ 1");
or:
if (arch_m64 == LP32 && arch_os == OS_CYGWIN)
add_pre_buffer("#define __CYGWIN32__ 1");
For the long term, I would prefer something like 4a) or 4b)
but currently it would only work for native builds.
So, I think that 3a) should be the best.
Best regards,
-- Luc
