PATCH: swapped usage of idx_t and bm_word_t types in bitMap.inline.hpp

[Dan Horák]

On Thu, 19 Sep 2013 14:14:18 +1000
David Holmes <david.holmes at oracle.com> wrote:

> On 17/09/2013 11:53 PM, Dan Horák wrote:
> > On Tue, 17 Sep 2013 22:48:46 +1000
> > David Holmes <david.holmes at oracle.com> wrote:
> >
> >> Hi Dan,
> >>
> >> On 17/09/2013 10:17 PM, Dan Horák wrote:
> >>> On Tue, 17 Sep 2013 12:39:22 +0200
> >>> Thomas Schatzl <thomas.schatzl at oracle.com> wrote:
> >>>
> >>>> Hi Bengt,
> >>>>
> >>>> On Tue, 2013-09-17 at 11:48 +0200, Bengt Rutisson wrote:
> >>>>> Hello Dan and Thomas,
> >>>>>
> >>>>> The change looks good to me to. Thanks for finding and fixing
> >>>>> this!
> >>>>>
> >>>>
> >>>> Thanks for the quick review.
> >>>>
> >>>> @Dan: The patch is now in the push queue and will go through the
> >>>> hotspot-gc mercurial tree.
> >>>
> >>> thank you guys, any action still needed from me?
> >>>
> >>> There are other places in the hotspot code where int and size_t
> >>> types are mixed (eg. in MAX2()/MIN2()) causing compile failures,
> >>> so I'll submit another patch for these issues.
> >>
> >> Have you actually eradicated all compile-time errors due to this
> >> issue? I worry that things will just unwravel as you try to do
> >> this.
> >
> > Some time ago the patch at [1] fixed all those conflicts and we
> > similar patch for openjdk7 in Fedora. But I guess it will need an
> > update for the recent sources.
> >
> > [1]
> > http://pkgs.fedoraproject.org/cgit/java-1.8.0-openjdk.git/tree/java-1.8.0-openjdk-size_t.patch
> 
> Shouldn't the integer promotion rules handle this without needing
> casts? I'm not clearly seeing the problem here.

IIRC the MIN2() and MAX2() are implemented as templates with base
integer types and this makes the problem when "size_t" (aka unsigned
long) and "unsigned int" go in as parameters


		Dan