Re: Please ensure SVGATextMode < 1.9-18 does not enter the distro for amd64
On Tue, Jul 31, 2007 at 12:18:46AM +0200, Luk Claes wrote:
> Kurt Roeckx wrote:
> > On Sat, Jul 28, 2007 at 05:11:36AM +0930, Ron wrote:
> >> Hi,
> >>
> >> Looking at the build logs for the latest upload, I see that
> >> un-blacklisting this for amd64 means the svgatextmode_1.9-15
> >> release (in testing) has been successfully built for amd64 now.
> >
> > It was actually build for stable, and uploaded to stable.
> >
> >> I don't believe we should distribute that one though, since the
> >> 'warning: large integer implicitly truncated to unsigned type'
> >> is a sign that outb() is being called with the port and value
> >> parameters reversed.
> >
> > This is something for the stable release team then. They should
> > probably reject svgatextmode_1.9-15 for etch.
>
> Package rejected.
>
> >> If we can keep anything prior to 1.9-18 still blocked that would
> >> probably be wise.
> >
> > I've set it to failed state in stable for now. Maybe the release team
> > will consider an upload to proposed updates to fix it?
>
> Well, it would be nice if we would know what it's about before considering
> it... Please tell us what's broken in the package in stable and how you want
> to fix it. Thanks already.
Most of the details should be recorded under #434068 now.
The short summary is:
- this has never been built for amd64 previously.
- we had a report from a user that it Works For Them.
- I don't see any inherent reason why it should not
and passed on the request to un-blacklist it.
- doing that didn't just build it for unstable, it also started
building older versions for stable too.
- actually building it myself revealed that without modification,
the amd64 builds would call outb() with the port number and value
swapped. This did not cause a build failure, but may cause any
manner of serious hardware failure if the resulting binary is used.
- The latest upload to unstable _should_ fix that, but because I don't
have the all the hardware needed to test this thoroughly, I still
consider it a 'blind' fix that needs more testing from other people.
So at this stage I think the best thing to do is to leave it out of
stable for now (for amd64, its fine for i386), until it has been better
tested and confirmed to work for at least several common video cards.
If we get that confirmation, then including a later build for amd64 in
stable should be a reasonably safe thing to do. This package has
changed very little over the last 4 - 5 years and is fairly solid on
x86, its just more or less a complete unknown on amd64 at this stage.
None of the changes I've made should affect the x86 builds, but they
also don't fix any bugs for that arch, so a proposed update probably
only makes sense if we do want to add it to the amd64 stable distro.
I'll look at that if we get a reasonable quorum of acks from people
who've actually tried it there without problem. Until then there
might be a few more things that shake out in 64 bit builds...
Did I miss anything?
Cheers,
Ron
Reply to: