wgrant | Well, it's damn slow, but I can see the errors now. | 00:03 |
---|---|---|
* wgrant stabs gcc with something very sharp. | 01:59 | |
wgrant | It seems to have decided that it can reorganise the request struct in different ways depending on whether it's on the client or server. | 02:00 |
=== johanbr_ is now known as johanbr | ||
jcristau | wgrant: ugh. | 10:12 |
wgrant | jcristau: gdb shows that the order of the struct on amd64 is precisely as it is sent over the wire. ie. wrong. | 10:13 |
jcristau | sounds like a serious gcc bug | 10:14 |
wgrant | Hmm. | 10:14 |
wgrant | I guess it's a bitfield, though. | 10:14 |
wgrant | So the compiler can do what it wants. | 10:14 |
wgrant | But I'm stumped as to why it wouldn't affect the other similar structs. | 10:14 |
wgrant | If it wasn't a bitfield, I believe that would be breaking the C standard (IIRC compilers can pack, but not reorder, struct members). | 10:15 |
wgrant | jcristau: Do you have hardware on which to reproduce it? | 10:16 |
wgrant | My ad-hoc crash course in X internals might have failed to reveal some critical piece of info. | 10:17 |
jcristau | i don't have an amd64 machine. debian does, so i could use that. | 10:19 |
wgrant | Yep, I used an Ubuntu-related machine. | 10:20 |
jcristau | you sure they're bitfields? | 10:22 |
jcristau | B16 and B32 seem to be defined to nothing if !WORD64 | 10:23 |
wgrant | Oh. Crap. | 10:23 |
mnemo | what prefix should I use when I install the libdrm bits from here http://www.intellinuxgraphics.org/download.html (the libdrm bits from this repository specifically --> git-clone git://anongit.freedesktop.org/git/mesa/drm | 11:36 |
mnemo | ) | 11:36 |
mnemo | i looked at the debian/rules file for the package libdrm2 and it says "/usr" | 11:37 |
mnemo | but on the other hand, if I use gdb's "info sharedlibrary" command on my stable hardy X I can see that X currently has /usr/local/lib/libdrm.so.2 loaded?? | 11:38 |
jcristau | mnemo: means you installed libdrm in /usr/local previously. don't do that. | 11:39 |
mnemo | okay, so if I delete the libdrm versino from /usr/local/lib and reboot it's likely that once I reconnect gdb I will see that it's using the correct libdrm then? | 11:41 |
mnemo | jcristau: does debian have some general rule/guideline/philosophy for when to use /usr versus /usr/local ?? | 11:43 |
jcristau | mnemo: /usr is for packages, /usr/local for locally installed stuff | 11:53 |
mnemo | /usr/local for stuff that the user installs himself from tarballs? | 11:54 |
Ng | so the lack of scrollwheel emulation I mentioned is only happening with 2.6.27-7 after the first suspend | 17:28 |
Ng | drop back to -6 and it works after multiple suspends | 17:28 |
Ng | with -7 once I've suspended, it just stops working | 17:29 |
wgrant | Ng: Aha, great. the changes aren't too huge to look through. | 21:22 |
wgrant | Oh. | 21:22 |
wgrant | Damn. | 21:22 |
wgrant | There're upstream changes too. | 21:22 |
wgrant | Forgot that. | 21:22 |
wgrant | Or maybe it does actually list all of them.. | 21:23 |
Ng | hrm | 22:02 |
Ng | wgrant: fwiw I filed it as bug 282387 | 22:54 |
ubottu | Launchpad bug 282387 in xserver-xorg-input-evdev "scrollwheel emulation breaks after suspend with 2.6.27-7" [Undecided,New] https://launchpad.net/bugs/282387 | 22:54 |
wgrant | Users are still being hit by the -evdev/-synaptics/g-c-c/g-s-d/nautilus removals. | 23:32 |
wgrant | Remarkable. | 23:32 |
wgrant | Both that their mirrors are so out of date, and that they don't notice. | 23:32 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!