From: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
---|---|
To: | ghaverla(at)freenet(dot)edmonton(dot)ab(dot)ca |
Cc: | pgsql-novice(at)postgresql(dot)org |
Subject: | Re: Third email on compilining 7.0.2 on Solaris 2.5.1 |
Date: | 2000-08-19 02:04:25 |
Message-ID: | 21042.966650665@sss.pgh.pa.us |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-novice |
ghaverla(at)freenet(dot)edmonton(dot)ab(dot)ca writes:
> I added a line to config.h in the section on vsnprintf
> which did a #include <varargs.h>
> This got past the initial compile problem I was having.
> How other architectures can include a vsnprintf() function
> without including varargs.h I don't know. Anyway, the
> compile goes quite a ways and dies, again on a varargs
> problem. A module assumes va_start takes 2 arguments,
> when my header files say only 1.
ANSI C says va_start() takes 2 args. <varargs.h> (on those platforms
that have it at all) defines an older, non-ANSI-compliant version of
the va_foo macros. What you want to be including is <stdarg.h>.
Dunno why src/include/c.h is not including that automatically on your
platform, but that's what to look at.
I've suspected for some time that the conditional include of <varargs.h>
near the bottom of c.h is dead code, if not actively pernicious. But
without access to a platform where
#if defined(sun) && defined(__sparc__) && !defined(__SVR4)
applies, I can't be sure whether to rip it out or not.
regards, tom lane
From | Date | Subject | |
---|---|---|---|
Next Message | ghaverla | 2000-08-19 04:27:41 | Re: Third email on compilining 7.0.2 on Solaris 2.5.1 |
Previous Message | Tom Lane | 2000-08-19 00:18:40 | Re: Installation Problem |