[PATCH 0 of 1] Enable building on MSYS/Cygwin shell using nativeGCC when configure executed via abs path.

[Ray Donnelly]

Any version of MSYS will do, provided the native version and the MSYS
version of the srcdir differ.

uname -a for me gives:
MINGW32_NT-6.1 computername 1.0.17(0.48/3/2) 2011-04-24 23:39 i686 Msys

.. but I've also had this problem on MSYS2 Alpha.

The only ways I can see to avoid this issue are:

1. Have srcdir only contain relative paths (e.g. configuring with
./configure rather than <ABS_DIR>/configure)
2. Unify somehow the layouts of MSYS and native Windows; one way to do
this would be to install MSYS to C:/usr and have the source live under
that, other ways include using "mklink /D" and MSYS's /etc/fstab
(test this by cd $SRCDIR; pwd -W and see if the output matches the
value of $SRCDIR, excluding the dirname difference.)
3. Have gmp-h.in and gmp-impl.h in a folder that, when prepended with
$srcdir, exists in your GCC include search paths.

.. anything else (and my testcase above) should fail. Can you please
try my testcase and post your config.log?

Best regards,

Ray.

On Tue, Jul 16, 2013 at 11:10 AM, Sisyphus <sisyphus at dodo.com.au> wrote:
>
> ----- Original Message -----
> From: "Ray Donnelly" <mingw.android at gmail.com>
>
>>I am sending it now as an attachment and crossing my fingers. I hope this
> is ok?
>>
>>> Which namespace differences?
>>
>> The bug the patch addresses is this one:
>> https://gmplib.org/list-archives/gmp-bugs/2013-March/003024.html
>
> But when *I*  build gmp in the MSYS shell (using MinGW), I don't experience
> any problems at all. It all works fine without any need for any patching.
>
> So ... the puzzle is "Why do *you* need this patch ?".
> What version of MSYS do you have ?
> According to 'uname -a' I'm using version 1.0.17 .
>
> Cheers,
> Rob
>
>