Re: [Guile-commits] GNU Guile branch, wip-1-8-mingw-build, updated. release_1-8-7-22-g7edad77

[Neil Jerram]

Neil Jerram <address@hidden> writes:

> address@hidden (Ludovic Courtès) writes:
>
>> Hi Neil,
>>
>> "Neil Jerram" <address@hidden> writes:
>>
>>> commit f41529c37b4a740aa555741ca78faa55fafcb3c0
>>> Author: Neil Jerram <address@hidden>
>>> Date:   Tue Mar 2 23:23:52 2010 +0000
>>>
>>>     Make test scripts check GUILE environment variable
>>>     
>>>     So that they use the guile (or guile.exe, in the MinGW case) that was
>>>     just built, and not just any guile that can be found in the PATH.
>>
>> The ‘TESTS_ENVIRONMENT’ variable in ‘test-suite/standalone/Makefile.am’
>> normally ensures that tests are run with the just-built guile, when run
>> with ‘make check’.
>
> Thanks.  I'll look into this more carefully then.

FWIW, the problem was the ".exe" extension that the MinGW cross-build
adds to generated executables.  On Linux, even if you have Wine
installed, so that NAME.exe is runnable, and even if NAME.exe is in the
PATH, the shell won't find NAME.exe if you ask it for NAME.

(When compiling on Windows I would guess that there isn't this problem.
CMD.EXE will certainly find NAME.exe, and I can imagine that the MSYS
shell has been patched to do that too.)

But anyway, following more careful thinking and feedback from the
"Reconsideration of MinGW" thread, I think that that method of
cross-building (i.e. with a cross-compiler and Wine on Linux) is pretty
unfruitful and I don't propose to continue it.

I'll leave the wip-1-8-mingw-build branch in the repository, in case
it's of interest, and because I don't know if branches can be deleted.
Is there a nice way of marking a branch as obsolete/historical, or of
adding an explanatory note to it?

Regards,
     Neil