**** BEGIN LOGGING AT Tue Mar 16 02:59:56 2010 Mar 16 14:36:05 NCommander: http://pastebin.mandriva.com/17521 Mar 16 14:36:13 dmart: ^ Mar 16 14:36:32 lool: ? Mar 16 14:36:36 dmart: Mandrive patch by rtp (Arnaud Patard) Mar 16 14:36:44 Which works for him Mar 16 14:37:04 It will work if the called function only has 4 args or less Mar 16 14:37:23 But if there are more, the saved lr will appear in the middle of them, messing things up. Mar 16 14:37:43 I don't know whether this framework is necessarily supposed to work for arbitrary numbers of args, but I'm assuming it is. Mar 16 14:37:55 dmart: I guess so as well, it's meant as a generic bridge Mar 16 14:38:03 I'm asking him to join #ubuntu-arm here Mar 16 14:38:13 Other than the issue of where on the stack lr is stored, the patch is effectively the same as mine, so it's reassuring that it works. Mar 16 14:38:17 dmart: I found what crashes valgrind on startup Mar 16 14:38:29 I have disappear to a meeting right now :( but I should be back in ~1hr or so Mar 16 14:38:47 rtp: dmart and NCommander are around and looked at your changes Mar 16 14:39:14 rtp: dmart will be back in an hour (meeting), but says the function probably has to work for any number of args Mar 16 14:39:26 lool: great. :) Mar 16 14:39:36 lool: ok. Mar 16 14:39:45 lool: I thought ARM already submitted a fixed version Mar 16 14:39:50 NCommander: did they? Mar 16 14:40:02 NCommander: where? Mar 16 14:40:13 NCommander: the patch in comment #69 doesn't work on my system Mar 16 14:40:16 lool: it was on the bug, give me a sec Mar 16 14:41:02 rtp: hrm, worked here, but that may be a fluke. Mar 16 14:41:16 rtp: lool: seems the patch never made it to LP, hold on, I have to go fishing for it Mar 16 14:41:44 NCommander: fwiw, I'm not running ubuntu on my systems :) Mar 16 14:42:27 rtp: lool: http://pastebin.ubuntu.com/395620/ Mar 16 14:42:31 rtp: The patch wasn't distro-specific : it ought work on Mandriva as well. Mar 16 14:42:47 rtp: of course not, but solving this bug should solve it everywhere. I know Debian, and Gentoo also were watching for a fix Mar 16 14:43:10 armin76: Did you test the patch? Mar 16 14:44:12 NCommander: will try this patch. Would bee nice to get it on the bug so that everyone can follow and test :) Mar 16 14:44:35 rtp: well it hasn't even been tested yet so lets not jump the gun so quickly ;-) Mar 16 14:44:45 rtp: I didn't even know Mandravia had an ARM port Mar 16 14:44:54 dmart: commented on the valgrind bug; happy if you have suggestions to avoid "mvn" on Thumb 2 Mar 16 14:45:18 NCommander: my intent was not to blame people at all. sorry Mar 16 14:45:44 rtp: no blame taken. I should have made sure the patch made it onto LP. That was just a breakdown in communications Mar 16 14:45:44 NCommander: our arm port is something new Mar 16 14:45:47 NCommander: even untested, the patch belongs in the bug report Mar 16 14:45:51 Where I've sentit now Mar 16 14:45:55 lool: thanks Mar 16 14:45:58 rtp: In the bug now, sorr Mar 16 14:46:01 y Mar 16 14:46:08 * lool & Mar 16 14:46:27 lool: thanks :) Mar 16 14:47:20 * NCommander hasn't run Mandriva since it was Mandrake Mar 16 14:51:04 well, at least, you know that Mandriva was Mandrake before Mar 16 15:03:25 NCommander: your pastebin patch is working here. Mar 16 15:04:03 rtp: ARM will be thrilled to know that. Do you have upstream commit access to ooo-build Mar 16 15:05:05 NCommander: no, I've no access to it. Maybe the debian ooo guy as commit access ? Mar 16 15:05:11 s/as/has/ Mar 16 15:05:23 rtp: I have someone I can poke to commit it :-) Mar 16 15:05:48 :) Mar 16 15:06:27 rtp: can you just comment that on LP? I'll nudge ccheney to replace the patch in ooo-build later today Mar 16 15:07:19 NCommander: I've no account on LP :) Mar 16 15:08:01 rtp: Are you sure? For a while LP was creating accounts for all sorts of open-source developer sorts (you may not know your password) Mar 16 15:08:14 rtp: Check at https://launchpad.net/people Mar 16 15:11:29 persia: it looks like it created something when I was "playing" with gnome-control-center. I'll prefer using my mail @work Mar 16 15:11:50 rtp: If you can log into that account, you can feed it a different address. Mar 16 15:12:57 ok. Mar 16 15:20:28 persia: looks like it worked. thanks Mar 16 15:20:53 rtp: No problem. Thanks for stopping by and helping out :) Mar 16 15:21:13 rtp: Feel free to ask if you run into anything with which we can help. Mar 16 15:31:07 persia: well, this bug was annoying a lot of people and needed to be fixed imho. I regret I was not able to look at it before the end of last week :( Mar 16 16:12:15 NCommander, rtp: Hi, I can discuss the OOo fix thing now Mar 16 16:16:22 lool: What was the background to the Valgrind issue? Mar 16 16:38:21 dmart: rtp tested the revised fix, and it works on mandriva Mar 16 16:39:33 rtp's own patch has the right idea, but appears incorrect in a couple of ways. Mar 16 16:42:08 dmart: I've not played with arm asm since some years so it's possible there are bugs. I've mostly played with mips asm theses days :) Mar 16 16:42:41 Fair enough Mar 16 16:44:47 The issues were a) you should keep the stack 8-byte aligned when calling functions (so you need to push an even number of regs, or adjust sp)... Mar 16 16:45:16 and b) I think the bridge is intended to work for functions with > 4 arguments. The way you push lr causes it to appear in the middle of the args, which is likely confuse functions with >4 args. My version just pushes lr last; otherwise the logic is much the same. Mar 16 16:51:48 I see... Mar 16 18:38:14 Is there any reason for lucid alpha3 to be relatively 'slow' out of the box compared to eg. angstrom for arm7? Mar 16 18:38:47 It feels rather sluggish, I do hope it's not march=armv-7? Mar 16 18:40:44 it is v7 and thumb2 by default Mar 16 18:41:08 on what HW do you see that ? beagle ? Mar 16 18:42:34 Another OMAP3430 board (Samsung H1 smartphone) Mar 16 18:43:02 (I borrowed the built rootfs from beagle) Mar 16 18:43:24 well, we dont see any sluggishness on any of our supported boards (actually quite the opposite) Mar 16 18:44:02 Odd Mar 16 18:44:18 so it might be a kernel feature you are missing that the lucid userspace relies on Mar 16 18:44:48 Also, is there a particular operation that is sluggish? Mar 16 18:46:27 Well, 'general' command line use using the serial/ethernet gadget (latter using sshd). Perhaps I should try to get some figures Mar 16 18:47:04 Have to admit I'm running root off sd, so perhaps angstrom used tmpfs for some bits or so. Mar 16 18:49:05 persia: which patch? Mar 16 18:50:25 armin76: The OOo patch for gcc 4.4 Mar 16 18:51:39 persia: nobody told me anything, so no Mar 16 18:51:57 but if it works for you there's no point me testing it :) Mar 16 18:52:49 armin76: Not for testing: I thought you might want it for application. Mar 16 18:53:13 If you had tested, that's a nice data point, but sharing is always good :) Mar 16 18:53:48 nah, i just tested it because NCommander told me to Mar 16 18:53:54 but thanks for the offer, i appreciate it :) Mar 16 20:11:49 http://blog.laptopmag.com/hands-on-marvell-armada-powered-smartbook-does-1080p-video Mar 16 20:11:51 interesting. Mar 16 20:12:04 Though, hinge that doesn't open to 180 degrees? bleh. Mar 16 21:02:43 rtp: sorry, thought I had pushed the patch, but the connection had been reset; thanks for adding it Mar 16 21:04:26 ok, seeing something unusual here during install... it appears that kernel messages are passing through to X, and I'm getting text artifacts on the screen from it Mar 16 21:04:46 grabbing a screencapture won't see it, and moving the mouse across the affected areas redraws it correctly Mar 16 21:04:51 rather annoying though Mar 16 21:05:12 anyone seeing that on the current lucid images? Mar 16 21:05:22 GrueMaster: ^ Mar 16 21:06:13 I haven't seen it yet on Dove, but I had another issue to attend to. Mar 16 22:16:40 weird... wine-ing "gpu caps viewer" reports a DIFFERENT version of MESA than native glxinfo reports! Mar 16 22:16:45 1.5 Mesa 7.7.1-DEVEL Mar 16 22:16:50 er, wrong tab. Mar 16 22:16:59 (a.k.a. wrong channel) **** ENDING LOGGING AT Wed Mar 17 02:59:57 2010