[cairo] target XLIB <-> Image

[Jacob =JoukJansen]

cworth at cworth.org wrote on 13-MAY-2005 02:28:17.63
>On Mon, 2 May 2005 09:41:37 +0200 (MET DST), Jacob (=Jouk) Jansen wrote:
>> cworth at cworth.org wrote on 30-APR-2005 03:10:07.30
>> >>  - All test tell me that they fail. The pictures with target Image look
>> >>    awfull,
>> >
>> >Could you describe "look awful" in more detail? Perhaps one of those
>> >pictures worth a thousand words would help? Otherwise this is "hard"
>> >to diagnose.
>> I placed the "fill-rule" png files on my FTP site
>
>I've attached them here as well for easier access for others, (the
>images are quite small).
>
>The results from the image backend are obviously very broken. But it's
>not obvious to me what might be going on here. Anyone have any
>theories based on looking at the images? And what are the specifics of
>your platform here (native integer width, endianness, etc.)?
I did run it on an Alpha chip with OpenVMS7.3-2
default integer width : 4 bytes
endianness            : little
float type            : IEEE (due to compilation with /float=ieee
single precision float : 4 bytes
C-compiler and run-time libs normally folow the POSIX standards.



>> >>            but the pictures with target XLIB look nice at first glance. In
>> >>    the latter case only pixels next to the "sharp" edges in the pictures
>> >>    are "wrong"
>
>Yes, those results look just fine. You've probably got an
>implementation of an older version of the Render extension than that
>with which the reference images were generated.
Thanks. Actually for those images I did not use the VMS machine as X-server.
I used an linux-FC3 machine as remote display via an ssh-connection.


>> >>  1) Am I right to conclude that the target XLIB test are OK and that the
>> >>     differences occur only due to the use of a different compiler and a
>> >>     different CPU chip?
>
>More likely just a difference in X server implementation.
OK

>> >>  2) Can you give me some hints to get the "target Image" tests better?
>> >>     i.e. what code should I try to debug first?
>
>
>For tracking this down, the place to start will be
>pixman_composite_trapezoids. The first thing to check is that the
>trapezoids look correct --- they should be because obviously the xlib
>backend is getting the right trapezoids.
I'll try to find some time. Since I do not see any calls in the packages
that are known problematic on OpenVMS (i.e. select (v)fork ioctl), my prime
suspicion is uninitilized variables. On OpenVMS not everything is
initialized to zero by default.

                    Jouk


Bush : All votes are equal but some votes are more equal than others.

>------------------------------------------------------------------------------<

  Jouk Jansen
		 
  joukj at hrem.stm.tudelft.nl

  Technische Universiteit Delft        tttttttttt  uu     uu  ddddddd
  Kavli Institute of Nanoscience       tttttttttt  uu     uu  dd    dd
  Nationaal centrum voor HREM              tt      uu     uu  dd     dd
  Rotterdamseweg 137                       tt      uu     uu  dd     dd
  2628 AL Delft                            tt      uu     uu  dd     dd
  Nederland                                tt      uu     uu  dd    dd
  tel. 31-15-2782272                       tt       uuuuuuu   ddddddd

>------------------------------------------------------------------------------<