Re: gtk-peer compile fixes for gcc-2.95

[Mark Wielaard]

Hi,

On Mon, 2004-04-05 at 00:49, Mark Wielaard wrote:
> > We set CFLAGS to -g in a specific case, that is sort of different
> > since it claims to work around bugs in the infamous gcc "2.96".  I'd
> > prefer we avoid this in general.
> 
> I have just removed this for now.
> Will test on a system with 2.96 when I can.

OK, finally had time to test it on gcc 3.3.3 and 2.95 (Debian GNU/Linux
unstable, 686) and gcc 3.2.2 and 2.96 (Red Hat 9, 686). Test results
were the same whether or not the native JNI libraries were compiled with
or without -O2. So I will now remove this CFLAGS clobbering.
Also had to remove the -std=c89 option to AM_CFLAGS because gcc-2.95 has
a bug when -ansi and -std=c89 are simultaneously defined. This should
not be a real issue though since -ansi implies ISO C90 (for 2.95, 2.96,
3.2.2 and 3.3.3 at least).

2004-04-11  Mark Wielaard  <address@hidden>

        * configure.ac (CFLAGS): Remove definition.
        (AM_CFLAGS): Remove -std=c89 option.

Cheers,

Mark

P.S. The flakiness of the [Server]SocketTests in Mauve make comparing
tests result much harder then it should be :{
 

diff -u -r1.16 configure.ac
--- configure.ac        9 Apr 2004 17:35:56 -0000       1.16
+++ configure.ac        11 Apr 2004 23:29:56 -0000
@@ -115,17 +115,12 @@
   dnl FIXME - does not allow cross compiling
   AC_C_BIGENDIAN
  
-  dnl We want to avoid -O2 at the moment, gcc 2.96 from Red Hat 7
-  dnl cannot compile some native functions (gcc errors) with this
-  dnl flag.
-  CFLAGS="-g"
-
   dnl We want warnings, lots of warnings :-)
   AM_CFLAGS=
   if test "x${GCC}" = xyes; then
     dnl We want ISO C90 pedantic ansi, but with longlong (jlong) support
     dnl and modern POSIX and BSD C library funtions/prototypes.
-    AM_CFLAGS='-ansi -std=c89 -pedantic -Wall -Wno-long-long -D_BSD_SOURCE'
+    AM_CFLAGS='-ansi -pedantic -Wall -Wno-long-long -D_BSD_SOURCE'
   fi
   AC_SUBST(AM_CFLAGS)

signature.asc
Description: This is a digitally signed message part