Tue 11 Jun 2013 03:54:23 PM EDT Jun 11 15:54:23 <dgilmore>	halfline: there is no desktop spin for arm because gnome doesnt run
Tue 11 Jun 2013 03:54:38 PM EDT Jun 11 15:54:38 <halfline>	dgilmore: hey
Tue 11 Jun 2013 03:55:07 PM EDT Jun 11 15:55:07 <halfline>	dgilmore: so llvmpipe breaks on arm? or was it something else?
Tue 11 Jun 2013 03:55:11 PM EDT Jun 11 15:55:11 <halfline>	(just curious)
Tue 11 Jun 2013 03:55:15 PM EDT Jun 11 15:55:15 <dgilmore>	im assuming thats what you were after me for reading back
Tue 11 Jun 2013 03:55:31 PM EDT Jun 11 15:55:31 <dgilmore>	halfline: ive not looked into it honestly
Tue 11 Jun 2013 03:55:31 PM EDT Jun 11 15:55:31 <halfline>	you pretty much got where i was going
Tue 11 Jun 2013 03:55:38 PM EDT Jun 11 15:55:38 <halfline>	was going to ask if you could try llvmpipe with it
Tue 11 Jun 2013 03:55:51 PM EDT Jun 11 15:55:51 <dgilmore>	halfline: happy to try
Tue 11 Jun 2013 03:56:01 PM EDT Jun 11 15:56:01 <halfline>	cool, thanks!
Tue 11 Jun 2013 03:56:15 PM EDT Jun 11 15:56:15 <dgilmore>	anything special i should need to do
Tue 11 Jun 2013 03:56:36 PM EDT Jun 11 15:56:36 <dgilmore>	halfline: mostly fbdev is the X driver in use :(
Tue 11 Jun 2013 03:56:47 PM EDT Jun 11 15:56:47 <halfline>	yum install @gnome-desktop mesa-dri-drivers should be enough
Tue 11 Jun 2013 03:57:47 PM EDT Jun 11 15:57:47 <dgilmore>	halfline: llvm is not entirely sane on arm
Tue 11 Jun 2013 03:58:02 PM EDT Jun 11 15:58:02 <dgilmore>	it makes a lot of assumtions that are debian specific
Tue 11 Jun 2013 03:58:31 PM EDT Jun 11 15:58:31 <ajax>	most of that is only relevant to clang, not llvm, afaik
Tue 11 Jun 2013 03:58:53 PM EDT Jun 11 15:58:53 <dgilmore>	ajax: mostly
Tue 11 Jun 2013 03:59:08 PM EDT Jun 11 15:59:08 <halfline>	dgilmore: a quicker test might be yum install glx-utils mesa-dri-drivers ; and then seeing if glxgears works
Tue 11 Jun 2013 03:59:54 PM EDT Jun 11 15:59:54 <dgilmore>	halfline: i have a local mirror and installing @gnome-desktop already
Tue 11 Jun 2013 04:00:04 PM EDT Jun 11 16:00:04 <halfline>	ah okay, cool
Tue 11 Jun 2013 04:05:12 PM EDT Jun 11 16:05:12 <dgilmore>	halfline: i know that rob clark had gnome running on his nexus 4 using the freedreno driver
Tue 11 Jun 2013 04:05:30 PM EDT Jun 11 16:05:30 <dgilmore>	but thats not a use case that i think we want to support
Tue 11 Jun 2013 04:05:44 PM EDT Jun 11 16:05:44 <halfline>	right
Tue 11 Jun 2013 04:06:10 PM EDT Jun 11 16:06:10 <halfline>	i think he's planning on doing a kms driver at some point
Tue 11 Jun 2013 04:06:14 PM EDT Jun 11 16:06:14 <dgilmore>	install onto sdcard is slow :(
Tue 11 Jun 2013 04:06:23 PM EDT Jun 11 16:06:23 <--	flash__ (ejabberd@pepelaz.jabber.od.ua) has left #fedora-devel
Tue 11 Jun 2013 04:06:24 PM EDT Jun 11 16:06:24 <dgilmore>	i happened to have a test image booted
Tue 11 Jun 2013 04:10:52 PM EDT Jun 11 16:10:52 *	j_dulaney just broke anaconda; attempting to reproduce
Tue 11 Jun 2013 04:27:31 PM EDT Jun 11 16:27:31 <halfline>	dgilmore: any luck? or is it still chuggachugging ?
Tue 11 Jun 2013 04:27:46 PM EDT Jun 11 16:27:46 <dgilmore>	halfline: still installing. almost done
Tue 11 Jun 2013 04:28:16 PM EDT Jun 11 16:28:16 <dgilmore>	i shouldhave rebooted into sata disk
Tue 11 Jun 2013 04:35:38 PM EDT Jun 11 16:35:38 <halfline>	dgilmore: I gotta run for the day, but i'll check back tomorrow.  thanks for looking!
Tue 11 Jun 2013 04:37:35 PM EDT Jun 11 16:37:35 <dgilmore>	halfline: okay, np. my testing previously i got the oh no something has gone wrong screen, even if trying to use gdm

Wed 12 Jun 2013 09:01:41 AM EDT Jun 12 09:01:41 <dgilmore>	halfline: http://paste.fedoraproject.org/18121/12458137/
Wed 12 Jun 2013 09:02:43 AM EDT Jun 12 09:02:43 <halfline_laptop>	dgilmore: interesting
Wed 12 Jun 2013 09:03:27 AM EDT Jun 12 09:03:27 <halfline_laptop>	dgilmore: can you (gdb) print MCID ?
Wed 12 Jun 2013 09:03:37 AM EDT Jun 12 09:03:37 <ajax>	and OpIdx
Wed 12 Jun 2013 09:03:51 AM EDT Jun 12 09:03:51 <halfline_laptop>	yea
Wed 12 Jun 2013 09:03:54 AM EDT Jun 12 09:03:54 <dgilmore>	halfline_laptop: yeah i have it up in gdb still
Wed 12 Jun 2013 09:04:10 AM EDT Jun 12 09:04:10 <halfline_laptop>	would be good to see if OpIdx is out of bounds
Wed 12 Jun 2013 09:04:33 AM EDT Jun 12 09:04:33 <halfline_laptop>	or if the OpInfo item isn't filled in for some reason
Wed 12 Jun 2013 09:06:31 AM EDT Jun 12 09:06:31 <dgilmore>	halfline_laptop: Attempt to use a type name as an expression
Wed 12 Jun 2013 09:06:44 AM EDT Jun 12 09:06:44 <dgilmore>	thats all that was printed
Wed 12 Jun 2013 09:07:13 AM EDT Jun 12 09:07:13 <halfline_laptop>	oh must be some c++ ism. try (gdb) print MCID.OpInfo
Wed 12 Jun 2013 09:08:01 AM EDT Jun 12 09:08:01 <ajax>	not going to help?  MCID is both a variable and a class name, so i think that's what the c++ part of gdb will say when the variable's been optmized away
Wed 12 Jun 2013 09:08:18 AM EDT Jun 12 09:08:18 <ajax>	p MI.getDesc()
Wed 12 Jun 2013 09:08:37 AM EDT Jun 12 09:08:37 <halfline_laptop>	oh someone has the source file open in front of them
Wed 12 Jun 2013 09:09:09 AM EDT Jun 12 09:09:09 <ajax>	yeah, it's pretty cool how we have the source to everything
Wed 12 Jun 2013 09:09:18 AM EDT Jun 12 09:09:18 <ajax>	makes debugging way easier
Wed 12 Jun 2013 09:09:48 AM EDT Jun 12 09:09:48 *	mcepl just joined so he lost pearls of ajax's sacrasm :(
Wed 12 Jun 2013 09:10:07 AM EDT Jun 12 09:10:07 <halfline_laptop>	<ajax> not going to help?  MCID is both a variable and a class name, so i think that's what the c++ part of gdb will say when the variable's been optmized away
Wed 12 Jun 2013 09:10:07 AM EDT Jun 12 09:10:07 <halfline_laptop>	<ajax> p MI.getDesc()
Wed 12 Jun 2013 09:10:07 AM EDT Jun 12 09:10:07 <halfline_laptop>	<halfline_laptop> oh someone has the source file open in front of them
Wed 12 Jun 2013 09:10:13 AM EDT Jun 12 09:10:13 <mcepl>	ajax: flash or ndividia?
Wed 12 Jun 2013 09:10:20 AM EDT Jun 12 09:10:20 <ajax>	neither
Wed 12 Jun 2013 09:10:23 AM EDT Jun 12 09:10:23 <ajax>	llvm
Wed 12 Jun 2013 09:10:25 AM EDT Jun 12 09:10:25 <halfline_laptop>	LLVM on arm
Wed 12 Jun 2013 09:10:29 AM EDT Jun 12 09:10:29 <mcepl>	I see
Wed 12 Jun 2013 09:10:33 AM EDT Jun 12 09:10:33 <ajax>	i was in fact being serious
Wed 12 Jun 2013 09:10:41 AM EDT Jun 12 09:10:41 <halfline_laptop>	well a little tongue in cheek
Wed 12 Jun 2013 09:10:42 AM EDT Jun 12 09:10:42 <mcepl>	oh
Wed 12 Jun 2013 09:11:21 AM EDT Jun 12 09:11:21 <ajax>	the subtext was more "people don't use the source even if they have it, why do we even bother" than "god i hate closed source"
Wed 12 Jun 2013 09:11:25 AM EDT Jun 12 09:11:25 <halfline_laptop>	sarcasm was more his way of saying "yea, you could too!"
Wed 12 Jun 2013 09:11:58 AM EDT Jun 12 09:11:58 <ajax>	it always amazes me when people say "how does (open source component) do this thing"
Wed 12 Jun 2013 09:12:00 AM EDT Jun 12 09:12:00 <halfline_laptop>	ajax: of course you being the llvm maintainer means you had the source already cloned and prepped probably :-)
Wed 12 Jun 2013 09:12:29 AM EDT Jun 12 09:12:29 <ajax>	fair enough
Wed 12 Jun 2013 09:12:36 AM EDT Jun 12 09:12:36 <mcepl>	ah
Wed 12 Jun 2013 09:13:39 AM EDT Jun 12 09:13:39 *	mcepl hasn't heard "Use The Source Luke!"™ for a long time, that's true.
Wed 12 Jun 2013 09:17:38 AM EDT Jun 12 09:17:38 <halfline_laptop>	ajax: looks like it just needs a OpIdx != MCID.getNumOperands() && in the conditional ?
Wed 12 Jun 2013 09:18:26 AM EDT Jun 12 09:18:26 <halfline_laptop>	ajax: theory being OpIdx is getting incremented in the if block above and then going out of bounds
Wed 12 Jun 2013 09:18:55 AM EDT Jun 12 09:18:55 <halfline_laptop>	almost every other place that looks into the OpInfo table does a bounds check first
Wed 12 Jun 2013 09:20:05 AM EDT Jun 12 09:20:05 <halfline_laptop>	or maybe that whole check should be moved into the else
Wed 12 Jun 2013 09:20:45 AM EDT Jun 12 09:20:45 <halfline_laptop>	it'd be kind of weird for there to be three operands
Wed 12 Jun 2013 09:21:18 AM EDT Jun 12 09:21:18 <halfline_laptop>	and the only way to keep the operand count to two is to move it into the else { }
Wed 12 Jun 2013 09:22:34 AM EDT Jun 12 09:22:34 <--	mcepl (~mcepl@184.109.broadband10.iol.cz) has left #fedora-devel ("Odcházím")
Wed 12 Jun 2013 09:23:05 AM EDT Jun 12 09:23:05 <halfline_laptop>	interestingly enough if you do that, then MI.getOperand(OpIdx).isImm() could be written as MO2.isImm() which pairs pretty nicely with the if (MO2.isReg()) above it
Wed 12 Jun 2013 09:24:11 AM EDT Jun 12 09:24:11 <halfline_laptop>	dgilmore: still around?  can you do the gdb print ajax asked about?
Wed 12 Jun 2013 09:24:26 AM EDT Jun 12 09:24:26 <dgilmore>	halfline_laptop: yeah its running
Wed 12 Jun 2013 09:24:51 AM EDT Jun 12 09:24:51 <dgilmore>	everything got really slow in gdb when i installed all the debuginfo packages
Wed 12 Jun 2013 09:25:24 AM EDT Jun 12 09:25:24 <dgilmore>	top
Wed 12 Jun 2013 09:25:37 AM EDT Jun 12 09:25:37 <dgilmore>	wrong keyboard
Wed 12 Jun 2013 09:25:49 AM EDT Jun 12 09:25:49 <FranciscoD>	..
Wed 12 Jun 2013 09:26:58 AM EDT Jun 12 09:26:58 <dgilmore>	halfline_laptop: i need to run soon, take my daughter to daycare. i need to work from a coffee shop this morning, getting my cast off finally at noon
Wed 12 Jun 2013 09:27:08 AM EDT Jun 12 09:27:08 <dgilmore>	i will have my chromebook with me
Wed 12 Jun 2013 09:27:29 AM EDT Jun 12 09:27:29 <halfline_laptop>	okay, good luck with your cast!
Wed 12 Jun 2013 09:29:16 AM EDT Jun 12 09:29:16 <dgilmore>	it finally returned
Wed 12 Jun 2013 09:29:34 AM EDT Jun 12 09:29:34 <dgilmore>	No symbol "MI" in current context
Wed 12 Jun 2013 09:30:57 AM EDT Jun 12 09:30:57 <--	twaugh (~twaugh@cyberelk.plus.com) has left #fedora-devel
Wed 12 Jun 2013 09:32:19 AM EDT Jun 12 09:32:19 <ajax>	oh lord
Wed 12 Jun 2013 09:33:00 AM EDT Jun 12 09:33:00 <ajax>	right, because it inlined the entirety of that function into the caller, so gdb has no idea where anything is
Wed 12 Jun 2013 09:33:49 AM EDT Jun 12 09:33:49 <ajax>	but it's literally fine because llvm's written in such a professional easy-to-maintain language
Wed 12 Jun 2013 09:34:07 AM EDT Jun 12 09:34:07 <ajax>	shooting myself in the throat, brb