| |
| |
| |
|
Comments:
<0> so quiet in here
<1> too quiet
<2> hola
<3> ajax ?
<4> Jes_ pong
<3> ajax, had a look at that backtrace I sent?
<4> yeah, let me dig that up agai
<4> n
<4> i didn't see anything too obvious
<3> :(
<3> my cow-orker has an IRIX box with xhost+ so I can test it from here now
<3> ajax: but xlib code kinda makes my stomach do weird things
<4> woo 96 stack frames
<3> ajax: yeah it's awesome
<4> okay, so that's in the client-side software renderer. it looks like it's just trying to grab pixels from la-la land, but the w/y/width all look pretty sane
<4> so my guess is you've somehow lost track of the image storage
<3> sorry, but it has to pay it's membership fee for lala land before accessing
<3> only thing I lost track of are my marbles :)
<4> so this only happens when the app is displaying on irix?
<3> yep
<3> remote display to irix is where we see it
<5> Irix with xhost+? What could possibly go wrong?
<6> yet another example of two wrongs not making a right?
<3> well this is for testing only, the guy could ssh to my box and tunnel it over
<4> Jes_: what version of cairo is this with? the version in rawhide bears little resemblance...
<3> 1.0.2
<3> all I have here
<4> Jes_: can you send the output of running 'xdpyinfo' against the irix box?
<3> sent
<4> 8bpp root window? uh.
<4> so, cairo is known not to work at all with pseudocolor visuals
<3> is it 8bpp? wouldn't surprise me
<4> that's what the 'depth of root window' line says, yeah. many programs will inherit that if not told otherwise
<3> so what does this mean? GNOME just segfaults on 8bpp now?
<4> that's been my experience, yeah
<3> nice :(
<3> so you're saying that this is 'not a bug'?
<4> no one's really bothered to look into it, since, who has 8bpp anymore
<4> nah, i still think it's a bug
<4> the server is advertising plenty of visuals, so gtk should be smart enough to pick a sensible one
<3> any suggestions for what I should do here? mark it 'wont be fixed' or file a bug?
<3> I see
<4> but, that'd be what i'd check first. stick ethereal between the ia64 and the irix machines and see what visual gets selected
<4> if it's picking one of the pseudocolor visuals then cairo (or gtk, or something) needs to be fixed
<3> mind holding my hand here?
<4> sure
<3> let me see if I have ethereal
<3> now I do :)
<3> ajax, what do I do next then?
<4> launch it, start capturing listening on eth0 for tcp packets on port 6000
<4> and then DISPLAY=purzel:0 pygtk-demo
<3> first time with ethereal :)
<4> gah, the interface changed when they renamed it wireshark
<3> ajax, hmmm looks like ethereal is fscked :(
<3> it won't allow me to do a capture here and doesn't see any interfaces
<4> you do need to run it as root
<3> am doing that
<4> ooo...kay then
<3> ajax: all I can say is blame your main competitor :(
<4> does the demo app display at all?
<3> you mean pygtk?
<4> yeah
<3> it works if I display locally yes
<4> yeah, but when you display to irix. it should get at least as far as mapping a window and drawing something
<3> it draws a window yes
<3> if I run it in gdb it launches the window on the box then segfaults (and the window stays until I stop gdb)
<4> that'll work. do that, get to the segfault, switch to the irix box, open a termial, run 'xwininfo' and click on the window for the demo window
<4> demo app. thing.
<3> hang on need to coordinate that with cow-orker
<4> gah, except xwininfo only tells me visual cl*** and depth, not exact visual, but oh well good enough
<3> tcpdump can't do what you need instead of ethereal?
<4> sadly i've forgotten how to decode X protocol in hex
<3> oh dear
<4> ethereal has nice symbolic names for things
<3> ok
<3> in your mbox
<4> kablam. depth of 8.
<4> so, either gtk itself, or the pygtk bindings, should probably be taught to avoid that deoth
<4> also i believe Xsgi has options to control the list of available visuals; you might want to try configuring the irix X server to only expose 24bpp visuals and see if pygtk-demo runs successfully then
<4> that I can't help you with though
<3> going to be hard for me to do that - know nothing about it
<4> find someone who does, i guess
<4> i _think_ the man pages documented it but it's been a long time since i've touched an irix system
<3> well it's also an issue that the owner will be unhappy having me change it
<3> but thanks for the info
<7> ajax: Owen recently explained the visual-selection rationale to me.
<7> ajax: Originally, (when any visual would work), it was decided to make GTK+ use the default visual in order to respect the configuration of the X server.
<7> But now that some visuals will not work, he agrees it could make sense to have GTK+ avoid visuals that will break things. The route to make that happen would be to file a bug against GTK+ (if it doesn't exist yet).
<7> Jes_: See also this cairo bug about adding support for 8bpp pseudo-color (with some patches):
<7> https://bugs.freedesktop.org/show_bug.cgi?id=4945
<8> keithp: feel free to rename fonts-persian.conf to farsiweb-fonts.conf
<9> behdad: ah, sur
<8> keithp: would have made life a bit easier if you changed cache version to 3.
<9> behdad: yeah, for RH it might
<9> behdad: didn't need to for distros shipping the previous stable version :-)
<8> ...
<8> keithp: you forgot to add ppc64?
<9> I don't *have* ppc64
<8> or do you need the signature again?
<8> so you add what you have only? :)
<9> behdad: no, I add only what I get
<9> heh
<8> ok, filing a bug
<9> My actual plan is to push it to debian unstable and get back build breakage on all of the debian architectures
<8> that's basically what I did for Fedora
<8> keithp: fontconfig supports smallcaps now?
<10> woo cool
<8> keithp: so, couldn't you also move the timeout and blanks in separate conf files, leaving fonts.conf essentially empty? :)
<8> keithp: and also update the comment block in fonts.conf to replace local.conf with conf.d/
<9> I'm trying to avoid opening files gratuitously
<8> right...
<8> next will be caching parsed configuration ;)
<9> we'd have to stat the config files anyway
<8> keithp: I need to add a config file before 30-urw-fonts.conf
<8> keithp: does 300 work?
<8> don't think so
<9> behdad: 29.9 does
<8> :D
<9> probably should have picked numbers in the middle of the range instead of 0 and 9
Return to
#freedesktop
or
Go to some related
logs:
CreateRecordset error '80041820'
frostwire troubleshoot
centos vesa vga=791
wowmapview dapper
ypops configure for suse 10.1
postgrey whitelist spf
apt-get pdm ubuntu
#debian
libswigpy.so.0
#perl
|
|