Re: bug in image compositing in -xlib

[Benhur Stein]

On 2003-04-10 18:04:09 -0300 Alexander Malmberg <alexander@malmberg.org> wrote:
> This is a known issue with recent development versions of freetype. See
> http://wiki.gnustep.org/index.php/FTC_Image_Desc .
I think I'll downgrade freetype, then (or continue to use xlib until freetype 
2.2 arrives)

> > the variable viewIsFlipped cannot be trusted
>
> That's true, it shouldn't be used. Is there any particular reason why it
> hasn't been removed completely yet?
I do not know, but it made me loose quite some time trying to figure out
what was going on. After being sure that it's value was not correct,
I made a fast search for uses of it in all of back, and could only find
one use in windows backend, apart from accessors. In my opinion it should
be removed.
In fact, what killed me was that the gstate for one view (with allocGState 
called)
was not always the same when in XGGState, and one of them had the correct value
for the flag while the other one hadn't. Maybe this will show up again later...

> Looking at the ctm isn't a hack, it's the correct way of doing it. Of
> course, to be completely correct, you need to handle arbitrary ctm:s. :)
Maybe the ctm could have an isFlipped method :-)

> Anyway, I'm not very familiar with -xlib, but if it correctly handles
> all combinations of flipped and unflipped views (ie. results match those
> on OS), it should be correct enough to handle almost everything -gui
> itself will ever do.
It handled them correctly for me, with both methods (copy and composite).

Benhur