Re: Fix for PR/13358

[Jack Lloyd <lloyd at randombit dot net>]

Doh. The decimal constant was too large to be signed, which caused another
warning and made the test fail (probably should not have sent that off
without waiting for the testsuite to complete). Here is a fixed test that
passes with the patch I sent, fails otherwise. -Jack

On Fri, 23 Jan 2004, Jack Lloyd wrote:

> On Fri, 23 Jan 2004, Zack Weinberg wrote:
> 
> > Jack Lloyd <lloyd@randombit.net> writes:
> > 
> > > diff -u output is attached. Hopefully this is the correct format for gcc
> > > patches: applies with patch -p0 < pr13358.patch from toplevel gcc
> > > directory. This is based on Wednesday's snapshot but should be clean
> > > against the current 3.4 branch.
> > 
> > Please also provide a test case properly formatted for gcc's
> > testsuite.
> > 
> > zw
> > 
> 
> As a start here is one I've been putting into g++.dg/warn. It sets
> -Wno-long-long and tries to create three long long (decimal, hex, and
> octal) without a suffix.
> 
> Would it be appropriate to also write one for expected failures that a)
> sets -Wlong-long or b) doesn't set anything and assumes -Wlong-long is the
> default?
> 
> I'm not sure that (b) is the right thing, but I'm willing to go along with
> whatever is considered the best approach for testing.
> 
> -Jack
>

Attachment: Wlonglong-1.C
Description: Text document