confirure incorrectly 'detects' stdbool.h and sets HAVE_STDBOOL in .../source3/include/config.h. I had to get rid of it manualy: diff -u ./source3/include/config.h.orig ./source3/include/config.h --- ./source3/include/config.h.orig 2010-05-26 09:28:24.000000000 -0500 +++ ./source3/include/config.h 2010-05-26 09:38:35.000000000 -0500 @@ -2116,7 +2116,9 @@ #define HAVE_STDARG_H 1 /* Define to 1 if you have the <stdbool.h> header file. */ +#ifndef __TANDEM /* wrongly recognized? */ #define HAVE_STDBOOL_H 1 +#endif /* Define to 1 if you have the <stdint.h> header file. */ #define HAVE_STDINT_H 1 But I'm well aware thai this is not the proper fix... Nye, Jojo
Forget about it, I must have made some mistake, as an attempt to reproduce this failed miserably, 'configure' does detect the absence of stdbool.h and reflects that in config.h. Guess I missee a 'make clean' between two 'configure' runs and the 2nd one deteted stdbool.h in ...lib/replace. See also bug 7461, which I believe can also get closed (I'm verifying this at the moment). Bye, Jojo
do we actually require to run a make clean before each configure? Shouldn't libreplace's configure see that the found stdbool.h is its own one and not throw its that one away after it found it? :-)
Well, I was only guessing what might have gone wrong. I couldn't reproduce it later. That's why I closed the bug. But if you have ideas what went wrong and how to fix, I'd not be agaist it...
this is a problem we see on other machines, too. I also stumbed over it on IRIX and another machine recently. Metze or Jelmer - do you have an idea how to easily limit the search for header files to the system include directories only and prevent that libreplace finds its own generated header files?
I guess the issue is that we shouldn't have lib/replace in the include paths during configure - we weren't doing so in the past at least. Is there any reason why we need to do so now?