Re: branch-1_4 - platform recognition macro

[Andreas Büning]

Eric Blake wrote:

> According to Andreas Büning on 7/4/2006 12:58 PM:
> > Eric Blake wrote:

> With that recommendation, I'm checking in the following.  Could you please
> do a fresh CVS checkout and see if you can run the testsuite now?

Done. platform_ma fails because neither __unix__ nor __windows__ is defined.
syscmd() in 79.sysval starts an interactive (!) shell so I had to enter "exit"
to continue with the tests. This behaviour of system("") is documented.
I'm sorry to hesitate you over and over again with the same topic. :-(


Checking ./76.platform_ma
../doc/m4.texinfo:3280: Origin of test
./76.platform_ma: stdout mismatch
4c4
< 1
---
> 0
Checking ./77.syscmd
Checking ./78.esyscmd
Checking ./79.sysval
exit 0
Checking ./80.sysval
Checking ./81.maketemp
Checking ./82.errprint
Checking ./83.errprint
Checking ./84.m4exit
Checking ./85.m4exit
Checking ./86.incompatibi

Skipped checks were:
  ./80.sysval ./81.maketemp
Failed checks were:
  ./76.platform_ma:out


> One
> thing I still wonder about is your platform recognition macro, and the
> testsuite does have a test to ensure that exactly one such macro is
> defined.  Since we can't detect signals to syscmd, we should not be
> defining __unix__ on OS/2 (does your compiler even pre-define __unix__, or
> is it just __EMX__?).  But my understanding is that OS/2 does not really
> qualify as __windows__, so should I add a third platform macro, __emx__?

I don't know what's the meaning of a platform macro. Their is no macro like
unix (neither upper nor lower case nor with or without undescores). In 
principle,
__EMX__ identifies the libc, not the operating system but nowadays it's used
on OS/2 only.


Regards,
Andreas