Re: C++ PATCH: Fix sprintf format

[Mark Mitchell <mark at codesourcery dot com>]

On Tue, 2003-06-17 at 23:36, Eric Botcazou wrote:
> > Hopefully this patch, applied to mainline and branch, will make the
> > regression-checker happy...
> 
> I don't know if this is better on PowerPC, but bootstrap is still (and again) 
> broken on SPARC:

 Eric --

You're still seeing a "\n" in the output, which suggests that you didn't
pick up the fix that came with the email above.

In cp/mangle.c:mangle_conv_op_name_for_type you should have:

!   /* Create a unique name corresponding to TYPE.  */
!   sprintf (buffer, "operator %lu", 
  	   (unsigned long) htab_elements (conv_type_names));

Do you have that code?  Or do you still have a version with a \n at the
end of the string literal?

I'm running a SPARC bootstrap now to see if I can reproduce your
problem.

With respect to the DejaGNU issue, I cannot reproduce it.  On the 32-bit
SPARC system, the test does pass with -m64.

Your testsuite/lib/gcc-dg.exp should contain:

# Like check_conditional_xfail, but callable from a dg test.

proc dg-xfail-if { args } {
    set args [lreplace $args 0 0]
    global compiler_conditional_xfail_data
    set compiler_conditional_xfail_data $args
}

If it does, then we'll have to figure out what's going wrong on
SPARC64.  The first thing would be for you to send me the .log file
output for the simd-5.c test.

-- 
Mark Mitchell
CodeSourcery, LLC
mark@codesourcery.com