1. SCons
  2. Core
  3. SCons

Wiki

Clone wiki

SCons / BugParty / IrcLog2009-05-13

17:18:01  *      Jason_at_intel (n=[chatzill@bementil-116.illinois.prairieinet.net](mailto:chatzill@bementil-116.illinois.prairieinet.net)) has joined #scons 
17:27:59  *      [GregNoel](GregNoel) is no longer marked as being away 
17:28:16  *      garyo-home (n=[chatzill@209-6-158-38.c3-0.smr-ubr1.sbo-smr.ma.cable.rcn.com](mailto:chatzill@209-6-158-38.c3-0.smr-ubr1.sbo-smr.ma.cable.rcn.com)) has joined #scons 
17:30:22  *      stevenknight (n=[stevenkn@67.218.107.49](mailto:stevenkn@67.218.107.49)) has joined #scons 
17:30:30  *      stevenknight is now known as sgk_ 
17:30:40  <[GregNoel](GregNoel)>     Hi, guys 
17:30:49  <sgk_> hey 
17:31:07  <[GregNoel](GregNoel)>     right on time; looks like everybody's here 
17:31:19  <garyo-home>   hlo 
17:31:58  <garyo-home>   shall we dive in with 804? 
17:32:22  <sgk_> 804:  defer again? 
17:32:26  <sgk_> only half joking... 
17:32:26  <[GregNoel](GregNoel)>     I think we should tackle 804 and 2404 as a pair.  Even if it's not the same problem, they're both related to lazy actions; it makes sense for only one person to have to open the code. 
17:32:41  <sgk_> agree w/GN 
17:32:45  <sgk_> re: one person 
17:32:46  <garyo-home>   makes sense. 
17:33:59  <garyo-home>   I suggest for 804 and 2404 we assign to research/Bill; they're low pri so if he doesn't get to them soon, no problem. 
17:34:17  <sgk_> sounds good to me 
17:34:32  <[GregNoel](GregNoel)>     p3 or p4? 
17:34:37  <garyo-home>   p4 
17:34:38  <sgk_> although it goes a little against the idea of "research" being a triage-soon-for-reclassification bucket 
17:34:42  <sgk_> p4 
17:34:47  <sgk_> GN, you okay with it? 
17:34:49  <[GregNoel](GregNoel)>     done 
17:34:49  <sgk_> research p4 
17:34:52  <sgk_> done 
17:35:06  <[GregNoel](GregNoel)>     2409, consensus 
17:35:20  <[GregNoel](GregNoel)>     2406, consensus 
17:35:31  <[GregNoel](GregNoel)>     2408, consensus 
17:35:41  <sgk_> rockin'... 
17:36:03  <[GregNoel](GregNoel)>     2410, who, but otherwise consensus 
17:37:00  <garyo-home>   I've already got two from this spreadsheet, but this one is *so* easy... 
17:37:12  <[GregNoel](GregNoel)>     I guess I'll take it. 
17:37:16  <garyo-home>   thx 
17:37:25  <[GregNoel](GregNoel)>     done 
17:37:27  <sgk_> thank you... 
17:37:35  <[GregNoel](GregNoel)>     2410 
17:37:43  <sgk_> give it to me 
17:37:49  <sgk_> this and the next are a google guy 
17:37:54  <sgk_> i'll thump on him for filing lousy bug reports 
17:38:05  <garyo-home>   :-) 
17:38:07  <[GregNoel](GregNoel)>     I was thinking the same thing... 
17:38:29  <[GregNoel](GregNoel)>     ok, 2.x p3 sk, done 
17:38:40  <garyo-home>   agreed. 
17:38:43  <sgk_> no context to the diffs, no problem description...  sheesh, the people that company will stoop to hiring... 
17:38:55  <garyo-home>   lol 
17:39:00  <[GregNoel](GregNoel)>     {;-} 
17:39:25  <sgk_> 2413 
17:39:34  <garyo-home>   2414, xml output: wontfix, suggest posting on wiki 
17:39:35  <sgk_> er 
17:39:36  <sgk_> 2414 
17:39:47  <sgk_> i like that even better than future 
17:40:01  <[GregNoel](GregNoel)>     agreed 
17:40:02  <sgk_> seems like "xml outputter" is just too vague and generic 
17:40:18  <sgk_> done 
17:40:21  <[GregNoel](GregNoel)>     done 
17:40:40  <sgk_> 2415 
17:40:52  <garyo-home>   2.x/p2/ludwig 
17:41:10  <garyo-home>   or 1.3, either one 
17:41:15  <sgk_> ok with assigning to Ludwig, me as backup if he's AWOL? 
17:41:21  <[GregNoel](GregNoel)>     OK, but I'll say that he can reassign it to sk 
17:41:21  <garyo-home>   yes. 
17:41:21  *      bdbaddog (n=[chatzill@adsl-71-142-86-81.dsl.pltn13.pacbell.net](mailto:chatzill@adsl-71-142-86-81.dsl.pltn13.pacbell.net)) has joined #scons 
17:41:28  <sgk_> agreed 
17:41:33  <garyo-home>   Hi, Bill. 
17:41:37  <bdbaddog>     Hi 
17:41:39  <sgk_> hey bill 
17:41:46  <Jason_at_intel>       wow step away for a minute and you are all are almost done 
17:41:47  <sgk_> 2415:  done 
17:41:50  <sgk_> 2416 
17:42:06  <garyo-home>   Hi Jason, big crowd tonight! 
17:42:19  <Jason_at_intel>       I see .. that is good :-) 
17:42:29  <[GregNoel](GregNoel)>     I  don't think 2316 is lazy action; it's failure to substitute the target 
17:42:33  <garyo-home>   2416: guess I'll take it, at least I'll look at it. 
17:42:47  <garyo-home>   I know the subst code pretty well. 
17:42:42  <[GregNoel](GregNoel)>     research or anytime? 
17:42:54  <garyo-home>   research is a good idea. 
17:42:56  <[GregNoel](GregNoel)>     done 
17:43:09  <[GregNoel](GregNoel)>     2417: Gary, a return code of -1 means "command not found" (at least in this case) 
17:43:36  <[GregNoel](GregNoel)>     Same failure in packaging tests; I don't know why it can't find "rm". 
17:44:20  <[GregNoel](GregNoel)>     I wonder about a missing PATH 
17:43:46  <garyo-home>   Really? 
17:43:55  <sgk_> [GregNoel](GregNoel):  if zsh is well-behaved 
17:44:35  <sgk_> yeah, that packaging test failure is a real pain 
17:44:42  <sgk_> i took a quick look once but nothing obvious 
17:44:47  <sgk_> PATH (as reported by buildbot) looks good 
17:44:50  <garyo-home>   I use zsh daily on Mac, Linux and Windows.  I'll try it, but it'll work fine.  Give it to me for research, I'll get more info from the OP. 
17:45:02  <[GregNoel](GregNoel)>     done 
17:45:05  <sgk_> s/packaging test/packaging buildbot step/ 
17:45:17  <sgk_> done 
17:45:34  <sgk_> 2418:  me, +vs_revamp 
17:45:55  <[GregNoel](GregNoel)>     Really?  Sure. 
17:45:56  <sgk_> or maybe +[VisualStudio](VisualStudio), we're still using that keyword, right? 
17:46:37  <[GregNoel](GregNoel)>     I just put "v" in the box and Firefox finds the right one 
17:46:13  <garyo-home>   not +[VisualStudio](VisualStudio), that's for *building* VS solution files. 
17:46:19  <garyo-home>   (iirc) 
17:46:20  <sgk_> okay 
17:46:26  <sgk_> that sounds right 
17:46:46  <garyo-home>   :-/ 
17:46:53  <Jason_at_intel>       does this reproduce? 
17:47:07  <Jason_at_intel>       I just tested this.. and it works for me 
17:47:08  <garyo-home>   Jason: what, 2418? 
17:47:15  <Jason_at_intel>       yes 
17:47:23  <garyo-home>   hey, maybe it's already fixed then. 
17:47:26  <Jason_at_intel>       i assumed cache means that it woudl not rebuild 
17:48:02  <garyo-home>   it's more complex than that.  See the bug report. 
17:48:15  <Jason_at_intel>       ok 
17:48:02  <sgk_> aha 
17:48:08  <sgk_> light just went on 
17:48:23  <garyo-home>   sgk: ? 
17:48:36  <sgk_> i was misreading the problem 
17:48:59  <garyo-home>   Is it that they aren't sideeffects or something? 
17:49:19  <garyo-home>   or side effects don't get retrieved maybe? 
17:49:23  <Jason_at_intel>       ahhh 
17:48:58  <sgk_> yeah, it would be good to retrieve the .pdb from cache, too 
17:49:25  <sgk_> but retrieving multiple target files from [CacheDir](CacheDir) with the same "build signature" isn't supported right now 
17:50:14  <sgk_> so it would involve some design work to support 
17:50:18  <garyo-home>   ok, so maybe not +vs_revamp after all, but 3.x?  Your call... 
17:50:26  <sgk_> yeah, 3.x 
17:50:35  <[GregNoel](GregNoel)>     priority? 
17:50:50  <sgk_> p2 or p3 
17:51:01  <sgk_> it's one of those things that's grating because we *should* be smart about it 
17:51:08  <[GregNoel](GregNoel)>     p2 then 
17:51:10  <sgk_> but it's not clear how widespread a problem it really is in practice 
17:51:16  <garyo-home>   p3 then :-) 
17:51:27  <sgk_> :-) 
17:51:27  <[GregNoel](GregNoel)>     {;-} 
17:51:30  <garyo-home>   (I don't really care, just being snarky) 
17:51:42  <sgk_> go with p2 
17:51:46  <[GregNoel](GregNoel)>     done 
17:51:58  <[GregNoel](GregNoel)>     2319, consensue 
17:52:03  <[GregNoel](GregNoel)>     2420 
17:52:33  <sgk_> Rob, me as backup if he's AWOL 
17:52:34  <sgk_> 2.x p3 
17:53:08  <[GregNoel](GregNoel)>     done; I'll add you as cc 
17:53:14  <garyo-home>   sounds good 
17:53:26  <sgk_> 2421:  garyo++ 
17:53:32  <[GregNoel](GregNoel)>     ++ 
17:53:40  <garyo-home>   (well of course I broke it first...) 
17:53:50  <garyo-home>   but thx. 
17:53:54  <[GregNoel](GregNoel)>     (that's why I gave it to you) 
17:54:11  <[GregNoel](GregNoel)>     That's the issues; report on 1.3? 
17:54:37  <sgk_> i've still been out of it;  bill, yt?  any progress? 
17:54:51  <sgk_> bdbadog? 
17:54:55  <sgk_> bdbaddog? 
17:55:11  <bdbaddog>     sorry distracted by my wife.. ;) 
17:55:38  <garyo-home>   I have one 1.3 bug I've still been working on but it's much more complicated than I'd thought when I started it, 2048.  May need to get deferred. 
17:55:53  <bdbaddog>     o.k. so I'm behind on that, but looks like I need to shuffle some logic around to make it not too messy for the HOST_* and TARGET_* initialization. 
17:56:20  <Jason_at_intel>       does 1.3 get vs vs_revamp 
17:56:34  <sgk_> Jason_at_intel:  yes 
17:56:34  <bdbaddog>     Seems like that logic should be in the Platform/*.py code. 
17:56:35  <garyo-home>   That's the plan right now anyway. 
17:56:40  <sgk_> the pacing item is merging vs_revamp to trunk 
17:56:53  <Jason_at_intel>       gary: what about the extra vars in MSVS you complained about? 
17:57:12  <garyo-home>   btw, I'm using vs_revamp (latest trunk) successfully at work w/ VS2003 and 2005.  Thanks for some well-placed help, Jason. 
17:58:05  <Jason_at_intel>       no problem.. I have  new version almost done of this... should address the need to redo this code everytime we add a new version 
17:58:20  <Jason_at_intel>       I'll post when done 
17:58:37  <garyo-home>   more table-driven? 
17:59:03  <sgk_> bdbaddog:  any pieces where you could use help? 
17:59:20  <bdbaddog>     So the issue at hand with the HOST/TARGET variable initialization in Platform/*.py is that the Environment() isn't initialized prior to the environment being initialized (if I remember correctly) 
18:00:11  <bdbaddog>     And I wanted to bounce it off of at least 1 other person prior to coding it up. 
18:00:33  <sgk_> profitable to do it here?  or take it off line? 
18:02:47  <bdbaddog>     I'll bow to the wisdom of others. if there are other topics to discuss, then we can do it another time. 
18:03:08  <[GregNoel](GregNoel)>     Nothing but time tonight; Steven is pacer on the shuttle 
18:03:10  <sgk_> i think this is the main topic -- GN, you have anything else to go over? 
18:03:23  <[GregNoel](GregNoel)>     nope 
18:03:34  <sgk_> ~10-15 minutes to shuttle stop 
18:03:48  <garyo-home>   I'm happy to listen 
18:04:01  <bdbaddog>     Lemme find the code in question. 
18:04:09  *      vsmatck has quit (kubrick.freenode.net irc.freenode.net) 
18:04:09  <bdbaddog>     Do you all have vs_revamp checked out? 
18:04:24  <[GregNoel](GregNoel)>     One aside: For what it's worth, I've got an update of [PlatformToolConfig](PlatformToolConfig) that focuses on the platform configuration phase.  The tool part of it isn't anywhere near complete, though.  Should I push it over for your comments? 
18:04:27  <Jason_at_intel>       I am interest if nothing else to learn more of the insides of Scons 
18:04:53  <sgk_> i do 
18:04:57  <sgk_> updating... 
18:05:13  <Jason_at_intel>       updated 
18:05:14  <bdbaddog>     I can't remember off the top of my head where Platform is initialized. anyone? 
18:05:34  <[GregNoel](GregNoel)>     Platform/<ins>init</ins>.py 
18:06:59  <bdbaddog>     Scons.Platform.Platform() is called from somewhere. that's the where I'm looking for. 
18:07:25  <bdbaddog>     But if you look at Platform/<ins>init</ins>.py Platform() 
18:08:16  <bdbaddog>     you'll see it returns a [PlatformSpec](PlatformSpec), and then assigns the platform module.generated to the <ins>call</ins> 
18:08:58  <sgk_> ah 
18:09:00  <bdbaddog>     So then is the generated in win32.py the right place to populated the HOST_CPU, and HOST_OS 
18:09:09  <bdbaddog>     I mean generate() in win32.py 
18:09:47  <bdbaddog>     Also I wanted to move get_architecture() from Tools/MSCommon/arch.py to win32.py 
18:10:12  <sgk_> i think the generate() in win32.py 
18:10:23  <bdbaddog>     In which case the generate() for each platform should set the HOST_OS/CPU and TARGET_{OS|CPU} as well. 
18:10:24  <sgk_> (and in the other platform-speciific modules) 
18:10:33  <Jason_at_intel>       HOST_ARCH? or HOST_CPU 
18:10:51  <bdbaddog>     I think we decided HOST_CPU to leverage autoconf/auto* nomenclature. 
18:11:05  <sgk_> yes, HOST_CPU for exactly that reason 
18:11:32  <Jason_at_intel>       I thought it was the other way.. as x86_64 is an architecture not a CPU 
18:11:40  <bdbaddog>     So does that sound like a reasonable reorganization of the code? 
18:11:47  <[GregNoel](GregNoel)>     (Actually, I argue it should set PLATFORM_{CPU,VENDOR,KERNEL,OS} since it could be different in different Environment()s.) 
18:11:48  <sgk_> why move get_architecture()?  win32.py implies an OS 
18:11:58  <sgk_> i was thinking our mapping was arch => cpu 
18:11:58  <bdbaddog>     Jason: Hmm it's a cpu family 
18:12:06  <bdbaddog>     get_architecture is os specific logic. 
18:12:14  <bdbaddog>     at least for win32, it's only for win32. 
18:12:38  <Jason_at_intel>       that is fine.. just worried about people wanting a AMD64 CPU or an INTEL64 CPU 
18:12:49  <sgk_> oh, right, because of looking for the magic PROCESSOR<ins>ARCHIT* variables 
18:12:53  <sgk_> okay, makes sense 
18:13:11  <sgk_> [GregNoel](GregNoel):  PLATFORM_* ? 
18:13:12  <bdbaddog>     plus the tools aren't initialized yet. 
18:13:17  <Jason_at_intel>       Greg: ya.. so I did not push the otehr two as i can't find a build use for them.. only a packing use.. I took what was safe 
18:13:23  <sgk_> we were converging on HOST_* and TARGET_* 
18:13:24  <Jason_at_intel>       not that it woudl not be added on later 
18:13:41  <garyo-home>   I like HOST and TARGET_*. 
18:13:42  <bdbaddog>     and that allows us to have (eventually) the platform logic set the default tools lists 
18:14:44  <sgk_> we're also converging on _{CPU,VENDOR,KERNEL,OS} suffixes to ride GNU's coattails 
18:14:54  <Jason_at_intel>       I thought HOST_ARCH was the "high level" cpu and CPU was the low level 
18:15:02  <[GregNoel](GregNoel)>     In general, whatever (cross-)compiler you want to invoke, it runs on the current platform, so you shouldn't really need the detailed specifics.  Only for what you're generating.  And PLATFORM_* makes sense for that. 
18:15:33  <sgk_> sorry, i don't understand that 
18:15:43  <Jason_at_intel>       which one? 
18:15:48  <sgk_> PLATFORM_ seems ambiguous to me 
18:15:49  <garyo-home>   google HOST_ARCH: 3960 results.  google HOST_CPU: 59,700 results. 
18:15:55  <sgk_> HOST_* and TARGET_* seem obvious 
18:16:07  <bdbaddog>     +1 HOST_ and TARGET_ 
18:16:23  <garyo-home>   +1 HOST_ and TARGET_ 
18:16:26  <Jason_at_intel>       but these don't map to auto config... i say HOST as Greg might say BUILD 
18:16:32  <sgk_> Jason_at_intel:  what would be the distnction between "high level" and "low level" ? 
18:16:33  <[GregNoel](GregNoel)>     I won't argue here; I'll push over the proposal; critique at your leisure. 
18:16:40  <sgk_> okay 
18:16:48  <sgk_> just reaching exit for shuttle stop 
18:16:57  <sgk_> < 1 minute 
18:17:14  <bdbaddog>     any reason not to start coding as proposed? 
18:17:19  <sgk_> Jason_at_intel:  examples of "high level" vs. "low level" ? 
18:17:20  <bdbaddog>     Naming aside ? 
18:17:23  <Jason_at_intel>       high level is x86, x86_64... low level is p3, p4, amd64 
18:17:28  <sgk_> and what it provides that the GNU model doesn't already cover? 
18:17:31  <garyo-home>   jason: I agree 
18:17:49  *      [BinkyTheClown](BinkyTheClown) (n=binky@unaffiliated/binkytheclown) has joined #scons 
18:17:52  <bdbaddog>     I think realisticaly the low level is left for the user to implement at this point. 
18:18:03  <garyo-home>   sgk: compiler flags to generate specific code (SSE2, SSE3).  Prob not that important for us. 
18:18:07  <bdbaddog>     it's flags on top of whatever flags are set for bit-ness 
18:18:15  <garyo-home>   bdbaddog: that's right, for now at least. 
18:18:19  <sgk_> iirc, boost distinguishes between 32-bit and 64-bit "memory model" 
18:18:27  <bdbaddog>     yes. not forever, but we have bigger fish to fry 
18:18:36  <Jason_at_intel>       I was was just simplifying term to a simple set, as i could not justify to other the need for the other stuff 
18:18:36  <sgk_> okay, shuttle 
18:18:41  <sgk_> good work, folks 
18:18:46  <sgk_> i'll check the log for follow-on discussion 
18:18:47  <bdbaddog>     l8r SGK 
18:18:50  <garyo-home>   ok, bye for now SGK 
18:18:50  *      sgk_ has quit (Read error: 104 (Connection reset by peer)) 
18:19:33  <bdbaddog>     Anyone have feedback + or - for my proposed reorg? 
18:19:56  <Jason_at_intel>       I think the move for get_architecture is correct 
18:19:58  <bdbaddog>     mainly focused on win32/visual studio/visual c initially. 
18:20:34  <bdbaddog>     I figure sunos/irix/hpux/etc would then handle their possible CPU's 
18:21:01  <bdbaddog>     in some sense OS===PLATFORM 
18:21:26  <garyo-home>   bdbaddog: seems OK to me. 
18:21:34  <[GregNoel](GregNoel)>     platform==unix, os==ultrix 
18:21:47  <bdbaddog>     :) ultrix 
18:21:51  <bdbaddog>     ahh memories. 
18:22:15  <[GregNoel](GregNoel)>     OK, os==solaris 
18:22:08  <Jason_at_intel>       does this mean we will say linux.. instead of posix? 
18:22:59  <[GregNoel](GregNoel)>     no idea.  you asked for an example. 
18:23:13  <garyo-home>   jason: yes, I'd say linux/bsd/irix, not just "unix" for all of them. 
18:23:27  <garyo-home>   .. or posix. 
18:23:38  <Jason_at_intel>       platform=linux os=RH 
18:23:53  <bdbaddog>     re posix; I agree, but not sure how much that might break in userland if we make that change. 
18:23:54  <Jason_at_intel>       or platform==os==linux 
18:23:58  <bdbaddog>     probably need deprecation cycle? 
18:24:04  <[GregNoel](GregNoel)>     Uh, that's vendor==redhat, os==gnu 
18:24:17  <bdbaddog>     os=GNU/Linux 
18:24:28  <garyo-home>   I would *not* say os=RH/Ubuntu; that's a distro, the OS is still linux. 
18:24:41  <bdbaddog>     anyway it's really semantics at some point. 
18:24:48  <[GregNoel](GregNoel)>     vendor==ubuntu 
18:24:55  <garyo-home>   Greg has it right there, vendor=ubuntu. 
18:24:56  <bdbaddog>     and none of that really impacts vs_revamp issues. 
18:24:58  <Jason_at_intel>       vender makes sence 
18:25:10  <garyo-home>   but it's not very relevant to Bill's task right now. 
18:25:14  <bdbaddog>     and none of that needs to happen in 1.3 
18:25:35  <bdbaddog>     we can add HOST_VENDOR/TARGET_VENDOR in 2.x 
18:25:45  <Jason_at_intel>       seem like a good idea 
18:25:52  <garyo-home>   sure, if it turns out to actually affect something :-) 
18:26:01  <bdbaddog>     and that leaves time to figure out which names we'll use. 
18:26:12  <garyo-home>   Actually it could, for packaging.  rpm vs. deb for instance. 
18:26:26  <Jason_at_intel>       and kernel drivers 
18:26:33  <bdbaddog>     true, but that's sort of user space issues. 
18:26:45  <garyo-home>   for sure. 
18:27:04  <bdbaddog>     if we add too much user space stuff to scons, we'll never improve it. 
18:27:11  <Jason_at_intel>       that is why i have not touched in yet in what i have worked on 
18:27:40  <bdbaddog>     and/or maybe in a contrib/unsupported/future package. 
18:27:48  <bdbaddog>     sorry module. 
18:28:19  <garyo-home>   contrib++ 
18:28:40  <bdbaddog>     O.k. so I'll start coding all that up (the HOST_OS|CPU, TARGET_OS|CPU) for win32. 
18:28:50  <garyo-home>   sounds great to me. 
18:28:54  <Jason_at_intel>       so is it _ARCH or _CPU 
18:28:59  <bdbaddog>     _CPU 
18:29:00  <Jason_at_intel>       i had this from our talk 
18:29:06  <Jason_at_intel>       <Jason_at_intel> 
18:29:08  <Jason_at_intel>       HOST/TARGET_OS _ARCH or ARCHITECTURE 
18:29:10  <Jason_at_intel>       <stevenknight> 
18:29:11  <Jason_at_intel>       i thought the consensus was that "platform" should conceptually mean the tuple of relevant things 
18:29:13  <Jason_at_intel>       <stevenknight> 
18:29:14  <Jason_at_intel>       _OS and _ARCH 
18:29:23  <Jason_at_intel>       I missed when this was changed 
18:29:43  <bdbaddog>     I think Steven and Greg conversed about this, and sugguested HOST_CPU 
18:29:50  <garyo-home>   Bill: I kind of like _ARCH myself for x86/x86_64/mips 
18:30:06  <bdbaddog>     I'm agnostic. 
18:30:11  <bdbaddog>     about _CPU or _ARCH 
18:30:20  <Jason_at_intel>       I was pushing for having CPU and ARCH 
18:30:29  <bdbaddog>     well. maybe not.. _ARCH is probably appropriate for this level of specificity 
18:30:30  <Jason_at_intel>       ideally add CPU later 
18:30:47  <bdbaddog>     Greg U still there? 
18:30:58  <garyo-home>   Yes, that's why I like arch.  Leaves room for more specific cpu. 
18:31:06  <[GregNoel](GregNoel)>     sorta; being distracted by pizza 
18:31:10  <bdbaddog>     :) 
18:31:46  <bdbaddog>     U have an opinion _CPU vs _ARCH when its at the level of x86 and x86_64 vs P4/Core2/Atom/ whatever 
18:31:54  <[GregNoel](GregNoel)>     I favor CPu because we can leverage autoconf stuff; if you deem it must be called arch, then that's not my problem. 
18:32:28  <Jason_at_intel>       how do you plan to leverage autoconf? 
18:32:40  <Jason_at_intel>       I thought you want to build in a autoconf like system? 
18:32:47  <bdbaddog>     we can default CPU = ARCH and then user/platform  can set/user more specific? 
18:32:49  <Jason_at_intel>       but not copy everything 
18:33:13  <garyo-home>   I think GNU uses ARCH for this, at least sometimes.  See [http://pingus77.free.fr/Gentoo/240/files/xdtv-2.4.0-mmx.patch](http://pingus77.free.fr/Gentoo/240/files/xdtv-2.4.0-mmx.patch) 
18:33:17  <garyo-home>   (which I just googled) 
18:33:22  <[GregNoel](GregNoel)>     Just about everybody who configures machines for cross-compiles knows GNU triples; I want people converting from Autoconf to be comfortable. 
18:33:30  <Jason_at_intel>       got to love google 
18:34:06  <Jason_at_intel>       i guess i never got autoconf to easy work for cross builds 
18:34:24  <Jason_at_intel>       it always was to hard to get it to work 
18:34:26  <garyo-home>   "OS: linux-gnu, ARCH: i386, CPU: i686" <--- from another google 
18:34:27  <[BinkyTheClown](BinkyTheClown)>        [GregNoel](GregNoel): that's me ;) 
18:35:01  <bdbaddog>     [GregNoel](GregNoel): I'm not that familiar with autoconf, what's their terms? 
18:35:25  <garyo-home>   OK [BinkyTheClown](BinkyTheClown): would you expect ARCH=x86 and CPU=Core2, or just CPU=x86? 
18:35:26  <[GregNoel](GregNoel)>     garyo-home, but what's the GNU triple?  It'll say x86. 
18:35:34  <Jason_at_intel>       I agree that we need at least  OS and ARCH for cross builds.. I don't see the need for vender or a CPU 
18:36:01  <Jason_at_intel>       I cross build all the time, but maybe I am missing something 
18:36:27  <[BinkyTheClown](BinkyTheClown)>        garyo-home: I'd expect CPU=Core2 
18:36:47  <Jason_at_intel>       the triple can also say intel64 amd64 and x86_64 
18:36:54  <[BinkyTheClown](BinkyTheClown)>        garyo-home: and ARCH=x86 
18:37:44  <[GregNoel](GregNoel)>     No, the triple can only say x86_64.  Anything else is canonicalized.  Try it. 
18:37:53  <Jason_at_intel>       so I am for the arch=x86 and cpu=p4r2-see2 
18:37:54  <garyo-home>   How do I try it? 
18:38:08  <[GregNoel](GregNoel)>     Look for config-sub. 
18:38:36  <[GregNoel](GregNoel)>     There's probably a copy on your system somewhere; if there's a configure.in, there's a config.sub. 
18:40:25  <garyo-home>   my config.sub (Ubuntu 9.04) allows x86, i386, i686, x86_64 at least for cpu 
18:40:27  <bdbaddog>     ok. so sounds like HOST_OS, HOST_ARCH (and future HOST_CPU) 
18:41:07  <Jason_at_intel>       gary: that is why i like _arch and _CPU 
18:41:37  <[GregNoel](GregNoel)>     try 'config.sub blah-garyo-linux' for blah in x86, i386, i686, x86_64, amd64, ... 
18:42:51  <garyo-home>   yup, I did.  It accepts (and returns exactly) those, except for amd64. 
18:43:17  <garyo-home>   (which it canonicalizes to x86_64). 
18:43:23  <garyo-home>   it's a shell script, btw. 
18:43:26  <[GregNoel](GregNoel)>     Then GNU considers them significantly different, and we should use them.  What more can I say? 
18:43:28  <Jason_at_intel>       I think the point we have to remember is why does the user care about these values 
18:44:22  <Jason_at_intel>       do 80% of user building care about i386 or i686.. I woudl say no 
18:44:33  <Jason_at_intel>       packaging they might care mroe.. but building .. no 
18:45:22  <garyo-home>   again, ARCH is high level (x86 vs. x86_64 vs. mips4 vs. powerpc), CPU should be lower level (386 vs 686 vs mmx) 
18:45:36  <[GregNoel](GregNoel)>     garyo-home: don't look in the shell script; it's contaminated with the GNU virus 
18:45:54  <garyo-home>   Oh no, I'm infected. 
18:46:26  <Jason_at_intel>       GNU virus? 
18:46:31  <[GregNoel](GregNoel)>     I've been very careful not to look inside, in case we have to reverse-engineer it. 
18:46:34  <garyo-home>   ah-choo! 
18:47:05  <bdbaddog>     o.k. I've gotta check out. But I'll code to HOST_OS, HOST_ARCH (with HOST_CPU to be future) 
18:47:08  <Jason_at_intel>       so everyone is saying  HOST_OS, HOST_ARCH (and future HOST_CPU) 
18:47:24  <Jason_at_intel>       is this agreeable? 
18:47:31  <bdbaddog>     I can also float it to the dev list. 
18:47:32  <garyo-home>   I think that will be pretty clear to everyone. 
18:47:38  <bdbaddog>     +1 
18:47:47  <Jason_at_intel>       +1 
18:47:50  <garyo-home>   bdbaddog: sure, mention it why not 
18:49:07  <bdbaddog>     O.k. Now to actaully do the work... ;) 
18:49:10  <bdbaddog>     Good night to all! 
18:49:13  <garyo-home>   Just for kicks, here's what "dpkg-architecture" says about this: 
18:49:17  <garyo-home>   garyo@server1:~$ dpkg-architecture 
18:49:19  <garyo-home>   DEB_BUILD_ARCH=i386 
18:49:21  <garyo-home>   DEB_BUILD_ARCH_OS=linux 
18:49:22  <garyo-home>   DEB_BUILD_ARCH_CPU=i386 
18:49:23  <garyo-home>   DEB_BUILD_GNU_CPU=i486 
18:49:25  <garyo-home>   DEB_BUILD_GNU_SYSTEM=linux-gnu 
18:49:26  <garyo-home>   DEB_BUILD_GNU_TYPE=i486-linux-gnu 
18:49:28  <garyo-home>   DEB_HOST_ARCH=i386 
18:49:29  <garyo-home>   DEB_HOST_ARCH_OS=linux 
18:49:31  <garyo-home>   DEB_HOST_ARCH_CPU=i386 
18:49:32  <garyo-home>   DEB_HOST_GNU_CPU=i486 
18:49:34  <garyo-home>   DEB_HOST_GNU_SYSTEM=linux-gnu 
18:49:36  <garyo-home>   DEB_HOST_GNU_TYPE=i486-linux-gnu 
18:49:37  <garyo-home>   and now to all a good night. 
18:49:45  <[GregNoel](GregNoel)>     G'day, mate. 
18:49:46  <Jason_at_intel>       night! 
18:50:02  *      bdbaddog has quit ("[ChatZilla](ChatZilla) 0.9.84 [Firefox 3.0.10/2009042315]") 
18:50:05  *      garyo-home has quit ("[ChatZilla](ChatZilla) 0.9.84 [Firefox 3.0.10/2009042316]") 
18:50:18  *      Jason_at_intel has quit ("[ChatZilla](ChatZilla) 0.9.84 [Firefox 3.0.7/2009021910]") 
18:50:18  *      [GregNoel](GregNoel) has been marked as being away 

Updated