Re: [PATCH, ira] Make ira printf type correct

[Vladimir Makarov <vmakarov at redhat dot com>]

Michael Meissner wrote:
> On Tue, Jun 10, 2008 at 01:32:16PM -0400, Jakub Jelinek wrote:
>   
> > On Tue, Jun 10, 2008 at 01:19:52PM -0400, Michael Meissner wrote:
> >     
> > > I tried to bootstrap the ira compiler on a 32-bit Red Hat system, and it did
> > > not bootstrap due to a type conflict between a fprintf that used %ld to print
> > > out types that are int.  This patch fixes the immediate problem, but perhaps it
> > > is better to cast the types to long before doing the multiplies in case of
> > > overflow.
> > >       
> > This will break 64-bit builds.  As we can't use %zd, because we still
> > support pre-ISO C99 hosts, I'm afraid casting both arguments to (long)
> > is the way to go.
> >
> >     
> > > 2008-06-10  Michael Meissner  <michael.meissner@amd.com>
> > >
> > > 	* ira-conflicts.c (build_conflict_bit_table): Make tracing fprintf
> > > 	type correct.
> > >
> > > Index: gcc/ira-conflicts.c
> > > ===================================================================
> > > --- gcc/ira-conflicts.c	(revision 136625)
> > > +++ gcc/ira-conflicts.c	(working copy)
> > > @@ -112,7 +112,7 @@ build_conflict_bit_table (void)
> > >    if (internal_flag_ira_verbose > 0 && ira_dump_file != NULL)
> > >      fprintf
> > >        (ira_dump_file,
> > > -       "+++Allocating %ld bytes for conflict table (uncompressed size %ld)\n",
> > > +       "+++Allocating %u bytes for conflict table (uncompressed size %u)\n",
> > >         allocated_words_num * sizeof (INT_TYPE),
> > >         allocno_set_words * allocnos_num * sizeof (INT_TYPE));
> > >    for (i = 0; i < max_point; i++)
> > >       
> > 	Jakub
> >
> >     
>
> Right.  I think I was thinking of a ILP64 system, and not LP64 where int is
> 64-bit also.  Here is a corrected patch:
>
>   
Mike, thanks for reporting the problem.  I've submitted a bit different 
patch.

http://gcc.gnu.org/ml/gcc-patches/2008-06/msg00607.html
> 2008-06-10  Michael Meissner  <michael.meissner@amd.com>
>
> 	* ira-conflicts.c (build_conflict_bit_table): Make tracing fprintf
> 	type correct.
>
> Index: gcc/ira-conflicts.c
> ===================================================================
> --- gcc/ira-conflicts.c	(revision 136625)
> +++ gcc/ira-conflicts.c	(working copy)
> @@ -112,9 +112,10 @@ build_conflict_bit_table (void)
>    if (internal_flag_ira_verbose > 0 && ira_dump_file != NULL)
>      fprintf
>        (ira_dump_file,
> -       "+++Allocating %ld bytes for conflict table (uncompressed size %ld)\n",
> -       allocated_words_num * sizeof (INT_TYPE),
> -       allocno_set_words * allocnos_num * sizeof (INT_TYPE));
> +       "+++Allocating %lu bytes for conflict table (uncompressed size %lu)\n",
> +       (unsigned long)allocated_words_num * sizeof (INT_TYPE),
> +       ((unsigned long)allocno_set_words * (unsigned long)allocnos_num
> +	* sizeof (INT_TYPE)));
>    for (i = 0; i < max_point; i++)
>      {
>        for (r = start_point_ranges[i]; r != NULL; r = r->start_next)
>
>