Known GLX Bugs
--------------


There is another file in this directory, Bugs-ToDo, some of which seems out 
of date, and other parts of which I'm not sure about. I'm going to start a 
new file here, which specifically lists the bugs I've come across and what I
am currently working on.


1) the glx protocol does not appear to work with Irix 6.2. I haven't been
able to test this. Is this specifically with Irix 6.2, or with any Irix?
Fixed.

2) I've upped glx to claim it is compliant with v1.2. This is not completely
true: It does not deal with ClientInfo correctly, although it does respond
to it. Added some to this. It now keeps track of the compatible version
shared by client and server. I need to use that version to allow/disallow
certain function calls.

3) Implement the rest of glx 1.2 features. 
update: Done.

4) There may be errors still the span writing function pointers in xsmesa2.c.
I've got everything correct that I can test (TRUDITHER, 5R6G5B, 8R8G8B8A),
but there are many cases I haven't been able to test. If your program seg
faults immediately upon execution, please email me and send the glx_debug.log
file that was output.
Should be fixed.

5) With the sample program Mesa/samples/blendeq, after initially rendering
the window, if I move it twice, the program crashes with the following
Xlib error: Xlib: sequence lost (0x10000 > 0x565) in reply type 0x0! I don't
know as much as I need to about X events and protocols, so this may take me
awhile to figure out. For the time being, I'm going to put this on low priority
unless it starts happening with other programs. New: there are also a series 
of hexidecimal numbers that should be drawn over the colored boxes in the 
middle, these do not appear.
update: now it core dumps..

6) With the sample program Mesa/samples/bitmap1, the 'GL' in the center is
improperly drawn in white, it should be red.
Update: it appears Mesa renders the 'GL' white the first time, then red every
time after that. I don't consider this a bug anymore.

7) The sample program Mesa/samples/oglinfo gives the following error:
Error: glXQueryExtensionsString() failed.
update: now it core dumps..

8) The sample program Mesa/book/model seems to draw some of the stenciled
lines wrong.
Fixed.

9) The sample program Mesa/book/polys draws the center and right image
completely wrong. I think this is highly related to error #8.
Fixed.

10) texprox dumps core on my machine. Interesting to note that the regular
texprox is unable to allocate a proxy. I'll have to look more into this.

11) sample program Mesa/demos/drawpix. Crashes on changing the pixel height
with the following X Error: X Error of failed request:  GLXBadRenderRequest
GL_UNPACK_SKIP_PIXELS appears to have an off-by-one bug that becomes more and
more obvious the further you scroll right.
Update: only appears to respond to arrow keys. No crashing. Weird. I don't see
the off-by-one bug anymore, but the image used to wrap around when it scrolled
off the right side of the screen, that doesn't happen anymore.
More: from an SGI client, the image is washed blue.

12) Sample Program Mesa/demos/gamma has bug 8/9.
Fixed.

13) AreTexturesResident doesn't work properly. I've worked and worked on this
one, but can't get it going. Nothing bad happens, it just returns the wrong
values.

14) There appears to be a problem with Normals and flat-shading. See book/accanti.
The sphere has a highlight off-center to the right and down. It should be to the
right and up. Many others demonstrate similar problems that I believe are related
(accpersp, sceneflat).
Update: I've tested this with SGI OpenGL. It has the highlight on the left side. 
For the time being, I'm going to assume the SGI is correct. The tested page will
indicate where on the sphere the highlight appears for all of these.
More: on sceneflat, the bad lighting appears to affect the cone and sphere, but not
the torus. I can't tell if this is just due to light placement, or if some objects
are really affected while others aren't. Each client's image appears the same whether 
the image is displayed on an SGI or x86 box, but is different depending on if the
client itself is an SGI or x86.

15) The image in book/image appears slightly misaligned when run from an SGI host.

16) book/texturesurf, the colors are all wrong.

17) book/surface, should be glossy?

18) DrawArrays, from an SGI host. From linux host, DrawArrays is faked and sent as
individual Vertices, Normals, etc. From an SGI host, the data comes out wrong. I've
been looking pretty hard at this one, and I'm wondering if the SGI host may be packing 
data wrong. 

19) Mipmapping. This is just being pointed out, I don't yet believe it's a bug. SGI's
OpenGL appears to have different perspective mappings than Mesa. The image comes out
looking a bit different. I don't consider this a bug (at least not for my glx), as the
image always looks the same on SGI no matter what client sends the data, and vice-versa
for Mesa.

Terence Ripperda
ripperda@engr.sgi.com



