[cairo] Pixman sampling coordinates

[Bertram Felgenhauer]

Carl Worth wrote:
> > Actually it's even true after my previous patch for  a1-traps-sample,
> > as long as pixman_fixed_t has at least 2*n fraction bits, which is
> > currently the case. But with the plans to change to 20.12 precision,
> > that may change soon.
> 
> Am I reading this correctly that your previous patch introduced this
> fragility, but with the new approach the fragility is gone?

Yes, that's what I was saying, although I now realize that I mixed
up my 'n's - n is N_BITS/2 here, so the fragility isn't as bad
as I thought. In fact, N_BITS bits of fraction are required for
the sampling to work correctly in any case.

I can write up the math if there's any interest. (It's simple, but
slightly messy.)

> I'm a little confused by what patch sequencing you intend here. Also,
> the patches you've attached, (which look like they came from "git
> show" or so), are not accepted by "git am". I think that's a bug in

Hmm, git am produces too many mails for my taste, with small patches
like these. I guess I could use git-format-patch and attach all the
files, but attachments in mailbox format also seem a bit silly.

> Do you have a complete git branch you can put somewhere that I can
> just pull?

I have now.
    git-pull git://people.freedesktop.org/~inte/pixman sampling

gitweb URL:
    http://gitweb.freedesktop.org/?p=users/inte/pixman.git;a=shortlog;h=sampling

> Let's please fix the above to keep the common code outside of the
> #ifdef, (perhaps a #else won't even be necessary). From this context
> it looks like an assignment to lx is missing from th #else case
> anyway, right?

Ok, I redid that patch. And you were right, there was a assignment
missing there, thanks.

Bertram