Re: Nasty bug in 2.6-beta10-hzoli10.2

[Zoltan Hidvegi <hzoli@xxxxxxxxxx>]

Vidiot wrote:
> 
> <I still suspect that your signames.h is wrongly generated.  Many systems has a
> <broken nawk which causes very similar problems.  Check that your signames.h
> <again.  It should begin like this (on SunOS 4.1.2):
> <
> </** signals.h                                 **/
> </** architecture-customized signals.h for zsh **/
> <
> <#define SIGCOUNT        31
> 
> Here is my file:
> 
> /** signals.h                                 **/
> /** architecture-customized signals.h for zsh **/
>  
> #define SIGCOUNT        0
>  
> #ifdef GLOBALS
>  
> char *sigmsg[SIGCOUNT+2] = {
>         "done",
>         NULL
> };
>  
> char *sigs[SIGCOUNT+4] = {
>         "EXIT",
>         "ZERR",
>         "DEBUG",
>         NULL
> };
>  
> #else
> extern char *sigs[SIGCOUNT+4],*sigmsg[SIGCOUNT+2];
> #endif
> 
> 
> 
> You are right in that it is wrong.  Now we have to figure out what is wrong
> with the awk script.  Why do I say that?  Because gawk has been used all
> along.  I've had it installed for ages and configure finds it.  Here is the
> line from the make:
> 
> 	gawk -f ./signames.awk < /dev/null > signames.h
> 
> Any ideas?

I think you forgot to remove config.cache before running configure and
configure used the cached place for signal.h, which was /dev/null.  Remove
config.cache and run configure again.  I really forgot to mention this.
While running configure, watch for the line:

checking where signal.h is located... /usr/include/sys/signal.h

If your configure finds something else, send me the output of configure and
the congig.log file.  In this case you can create the signames.h by hand using

gawk -f ./signames.awk < /usr/include/sys/signal.h > signames.h

Otherwise remove signames.h and do a make.

Zoltan