Re: [screen-devel] fix build failure on MacOS X

[Christian Ebert]

* Micah Cowan on Thursday, August 07, 2008 at 12:07:46 -0700
> Christian Ebert wrote:
>> Latest git source fail to compile here (MacOS 10.4.11).

oops, I should have shown how it chokes:

$ make
CPP="gcc -E -g -O2 -I/usr/local/include -I/sw/include" srcdir=. sh ./osdef.sh
AWK=gawk CC="gcc -g -O2 -I/usr/local/include -I/sw/include" srcdir=. sh 
./comm.sh
gcc -c -I. -I.    -g -O2 -I/usr/local/include -I/sw/include screen.c
In file included from screen.c:51:
config.h:569: error: parse error before ‘/’ token
In file included from /usr/include/locale.h:40,
                 from screen.c:75:
/usr/include/_locale.h:68: error: parse error before ‘}’ token
make: *** [screen.o] Error 1

>> diff -r 186c11906163 src/acconfig.h
>> --- a/src/acconfig.h Thu Aug 07 02:34:28 2008 -0700
>> +++ b/src/acconfig.h Thu Aug 07 15:16:43 2008 +0200
>> @@ -564,7 +564,6 @@
>>  * If you are on a SYS V machine that restricts filename length to 14 
>>  * characters, you may need to enforce that by setting NAME_MAX to 14
>>  */
>> -#undef NAME_MAX             /* KEEP_UNDEF_HERE override system value */
>> #undef NAME_MAX
>> 
>> /*
> 
> Hm. Strange that it didn't affect you before, when it was present in
> config.h.in (acconfig.h was simply "git mv"ed from config.h.in).

Yes. Strange.

> Also strange that it would produce any error at all, since it is
> perfectly legal to #undef the same name multiple times, according to the
> C standard (including C89). Are you really certain that was the source
> of your problem?

See above. Apparently it doesn't like the comment after #undef.

Sorry. I know less than nothing about this autoconf stuff. I just
reported how I made it work for me.

>> BTW, I see there was a switch from config.h in to acconfig.h; I
>> now get the following warning:
>> 
>> autoheader-2.62: WARNING: Using auxiliary files such as `acconfig.h', 
>> `config.h.bot'
>> autoheader-2.62: WARNING: and `config.h.top', to define templates for 
>> `config.h.in'
>> autoheader-2.62: WARNING: is deprecated and discouraged.
>> autoheader-2.62:
>> autoheader-2.62: WARNING: Using the third argument of `AC_DEFINE' and
>> autoheader-2.62: WARNING: `AC_DEFINE_UNQUOTED' allows one to define a 
>> template without
>> autoheader-2.62: WARNING: `acconfig.h':
>> autoheader-2.62:
>> autoheader-2.62: WARNING:   AC_DEFINE([NEED_FUNC_MAIN], 1,
>> autoheader-2.62:                [Define if a function `main' is needed.])
>> autoheader-2.62:
>> autoheader-2.62: WARNING: More sophisticated templates can also be produced, 
>> see the
>> autoheader-2.62: WARNING: documentation.
> 
> Yup. A big warning over a very small thing.
> 
> Switching to acconfig.h is still better than what we had (which also
> produced a series of warnings). The reason you may not have seen
> warnings is that previously, the configure script was also checked in,
> and so you probably didn't need to regenerate it. The configure script
> and Makefile have since been removed from the repository (they will,
> naturally, be available in release tarballs).
> 
> Using acconfig.h is a temporary measure, but at least now we can be sure
> that everything defined in configure.in has a template in config.h. The
> next step would be to replace acconfig.h with templates defined directly
> in configure.in (or, perhaps, an .m4 file included thereby).

I can live with the warning just fine. Just wanted to give
feedback as I saw that this was changed recently.

c
-- 
  Was heißt hier Dogma, ich bin Underdogma!
[ What the hell do you mean dogma, I am underdogma. ]

_F R E E_  _V I D E O S_  -->>  http://www.blacktrash.org/underdogma/