Clone wiki

SCons / BugParty / IrcLog2008-10-01

16:50:28 * garyo-home ( has joined #scons 17:58:21 * stevenknight (n=stevenkn@ has joined #scons 17:58:47 <stevenknight> hey all 17:59:09 <garyo-home> Hi Steven; I' 17:59:11 <garyo-home> m here 17:59:40 <garyo-home> I've been working on a SCons talk for end of next week. I'll share it when I'm done. 17:59:48 <stevenknight> cool 18:00:08 <stevenknight> i've got slides from about three talks that i need to make public 18:00:18 <stevenknight> iirc you asked me for some awhile back and i never got back to you... :-( 18:00:26 <garyo-home> true... 18:00:51 <garyo-home> if you have any you'd like to send my way feel free, though at this point i think I have enough (about 26 slides for an hour talk) 18:00:56 * GregNoel is no longer marked as being away 18:01:11 <GregNoel> Hi, I'm here... 18:01:16 <garyo-home> Hi, Greg. 18:01:21 <stevenknight> hi greg 18:01:26 <stevenknight> garyo: what's the audience? 18:02:02 <garyo-home> about 20 people, small company developers. It's a group of companies under the same investor umbrella. 18:02:17 <stevenknight> actually, we should probably start, i'm on the shuttle but only for ~30 more minutes 18:02:26 <stevenknight> garyo-home: so kind of a general intro? 18:02:27 <garyo-home> ok I'm ready. 18:02:39 <GregNoel> 2201 18:02:46 <garyo-home> yes, intro but technical. Focus on cross-platform. Yes, 2201. 18:03:12 <GregNoel> Steven says invalid, but I don't know his reasons. 18:03:00 <stevenknight> 2201: consensus invalid? jean can re-open if he comes up with a case 18:03:18 <GregNoel> ah 18:03:12 <garyo-home> yes. 18:03:22 <GregNoel> ok, done 18:03:32 <stevenknight> i was going from your comment, actually 18:03:34 <GregNoel> 2217: Gary, how does one 'wrap things up with subprocess' when it's a Python action? External commands can use subprocess, I expect, but I don't see how to do it with a Python action. 18:03:37 <stevenknight> 2217: consensus wontfix 18:04:14 <garyo-home> 2217: right, subproc would only work to fork external actions. No way to multiprocess pure python. 18:04:23 <garyo-home> Especially with chdir! 18:04:04 <stevenknight> i think you could do it by forking a process for each worker thread 18:04:25 <stevenknight> and sending some indication to "execute this python action" 18:04:35 <stevenknight> but it would be really hairy to work out the details 18:04:37 <stevenknight> and not worth it 18:04:41 <garyo-home> right, not worth it! 18:04:47 <GregNoel> concur 18:04:48 <stevenknight> agree re: especially chdir 18:04:51 <stevenknight> okay, done 18:05:00 <garyo-home> 2219, I did it 18:05:04 <stevenknight> 2219: garyo++ 18:05:11 <GregNoel> ditto 18:05:14 <garyo-home> thx 18:05:21 <garyo-home> too cool to pass up 18:05:21 <GregNoel> 2220 18:05:21 <GregNoel> Gary, unfortunately, Repository() won't map subdirectories; it only maps some directory to the top of the build. And I'm pretty sure it's not working as designed; it certainly surprised me. 18:05:21 <GregNoel> Steven, maybe I have a convert's fervor, but you did such a good job of convincing me, that I don't believe that Repository() is sufficient for all cases. The case of a variant in a repository is the killer; the search order is context-dependent. 18:05:21 <GregNoel> (If you want to propose some new mechanism, I'll be happy to shoot holes in it... {;-}) 18:05:21 <GregNoel> It's easy enough to characterize what happened: add a source/foo.c and try to build it as Object('generate/foo.c'). SCons removes it before trying to build it. That is, SCons is treating it like a target, even though there's no source in the 'include' directory. 18:05:21 <GregNoel> I can add the failing SConstruct to the issue or create a new issue if you want. 18:05:25 <garyo-home> 2220? 18:06:06 <stevenknight> Greg: cool, please add th failing SConstruct 18:06:10 <stevenknight> let me research 18:06:23 <GregNoel> OK, research, Steven? 18:06:25 <garyo-home> sounds like consensus to me 18:06:41 <stevenknight> the approach we've had success with on the Chromium (a.k.a. Google Chrome) build is to use addRepository() (internal method) 18:07:06 <stevenknight> to hook up "source" directories to arbitrary build directories 18:06:54 <garyo-home> Chrome is using scons? 18:07:05 <GregNoel> Why else would they hire him? {;-} 18:07:17 <stevenknight> yes, Chrome is using SCons 18:07:28 <stevenknight> although the current release version is still pure Visual Studio 18:07:25 <garyo-home> can I say that in my talk? 18:07:47 <stevenknight> yes, it's public knowledge (it's even in wikipedia!) 18:07:59 <stevenknight> (and they wouldn't let them put it there if it weren't true!) 18:08:08 <GregNoel> suuuuuure 18:08:11 <garyo-home> :-/ 18:08:24 <stevenknight> Google Earth is also built with SCons 18:08:31 <garyo-home> nice! 18:08:44 <garyo-home> ok, 2221? 18:08:55 <stevenknight> pretty much all Google client-side apps are or will be SCons 18:09:08 <stevenknight> ("client-side" == Google for "runs on the user's system, not on a server") 18:09:15 <garyo-home> yup, I got that. 18:09:24 <stevenknight> it threw me when i got here 18:09:19 <GregNoel> 2221, I see 2145 is already assigned to Steven for research; shall I assume that both of them should be set to 1.2 p3? (It would clear off one of your pending research issues, Steven...) 18:09:25 <stevenknight> anyway 18:09:33 <stevenknight> yes, 1.2 p3 18:09:44 <GregNoel> ok, done 18:10:06 <stevenknight> 2222: 18:10:17 <garyo-home> 2222: mark me as cc and I'll watch it. 18:10:25 <GregNoel> done 18:10:35 <stevenknight> done 18:10:39 <garyo-home> (I'll have to make sure my issue query finds issues w/ me as cc...) 18:11:14 <GregNoel> If it doesn't, the BugSchedule query does. 18:11:18 <garyo-home> 2223: seems like David was working on this on the ML. Side effects might be the way to go? 18:11:32 <GregNoel> 2223, consensus David, but when? 18:11:23 <stevenknight> 2223: 2.x, p2, david? 18:11:34 <garyo-home> I'm fine w/ that. 18:11:41 <GregNoel> done 18:11:43 <garyo-home> 2.x p2 fine w/ me 18:12:04 <GregNoel> 2224 18:12:09 <stevenknight> 2224: 1.2 p2 david 18:12:11 <garyo-home> 2224: grr I need to get my buildbot vm up. I'll try vs_revamp on another windows machine soon. 18:12:22 <garyo-home> yes, 1.2 p2 david. 18:12:27 <GregNoel> done 18:12:42 <GregNoel> And that's the current spreadsheet! 18:12:24 <stevenknight> you'll probably run into the problems i had 18:12:32 <stevenknight> and the reasons he withdrew it from 1.1 18:12:46 <stevenknight> the other tool modules like mslink still call old code to find and set versions 18:12:50 <stevenknight> and that gums up his stuff 18:12:56 <stevenknight> i don't think it'll be too bad to fix 18:13:11 <garyo-home> I'll see if I can help, I think it's important. 18:13:19 <stevenknight> agreed re: important 18:13:22 <garyo-home> Just need to get a clean windows box to work on. 18:13:35 <stevenknight> 1.1 is about to go release candidate, though, hence 1.2 18:13:42 <stevenknight> (should have RC'ed already...) 18:13:42 <garyo-home> yes, that's OK. 18:13:59 <stevenknight> on to 2005? 18:14:06 <stevenknight> as in 2005[hq]x? 18:14:09 <garyo-home> Too bad about my 1.1 issues... all will get deferred to 1.2. 18:14:15 <stevenknight> mine too 18:14:22 <garyo-home> I can't edit the 2005h2 ssheet, but others are fine. 18:14:33 <stevenknight> i had editing problems too 18:14:50 <stevenknight> i was able to edit 2005q1 18:15:25 <GregNoel> The "share" toggle wasn't selected, although I've checked on it before; I set it now for sure, so it should be OK 18:15:36 <garyo-home> I don't think I can edit any of them, I'll try again now. 18:15:53 <stevenknight> GregNoel: i did get in to 2005h2 shortly before the meeting, thanks 18:16:03 <GregNoel> make sure you follow the "edit" link first 18:17:04 <garyo-home> well, doesn't work for me yet, but that's OK for now, let's look at a few of them. 18:17:18 <stevenknight> i have less than 10 minutes (closer to five), since we haven't pre-edited how can we make best use of the time? 18:17:34 <garyo-home> we're doing h2, right? 18:17:39 <GregNoel> yes 18:17:42 <stevenknight> 2005h2 yes 18:17:54 <garyo-home> I agree w/ you guys on the first 2 18:18:08 <garyo-home> 1259, 1209 18:18:15 <stevenknight> okay, 1259: worksforme done 18:18:26 <stevenknight> 1209: dup 324, done 18:18:35 <GregNoel> Steven, should we move the schedule to Mondays? 18:18:44 <stevenknight> 1210: i agree w/research 18:18:47 <stevenknight> schedule: 18:18:55 <stevenknight> sorry, i should have weighed in 18:19:09 <stevenknight> day of the week is less an issue for me than time 18:19:14 <GregNoel> 1210, but who? 18:19:29 <stevenknight> i have to get on the shuttle ~17:15-17:30 PDT 18:19:36 <stevenknight> and off ~18:30-19:00 PDT 18:19:56 <stevenknight> i've started biking to/from the shuttle, so it's 30 minutes off-line 18:19:57 <GregNoel> I meant the release schedule: from Sat to Mon 18:20:04 <stevenknight> oh! 18:20:08 <stevenknight> heh 18:20:15 <stevenknight> yes, Monday for those 18:20:38 <GregNoel> OK, I'll update the roadmap page 18:20:51 <garyo-home> 1210: I think it should be low priority, but research means look into it now. I vote not research but 2.x or something 18:21:00 <stevenknight> 1210: i'd rather it be someone else 18:21:22 <stevenknight> i think it needs some thinking unencumbered by my knowledge of the Dir node code 18:21:39 <GregNoel> (We'll have to slip 1.1 by a few days) 18:21:47 <stevenknight> right re: 1.1 slip 18:21:59 <stevenknight> 1210: shall we see if bill can take it? 18:22:04 <garyo-home> Probably just look at the dir after each removal, and if dir is empty and under a build dir, remove dir. 18:22:12 <garyo-home> (after each FILE removal, sorry) 18:22:52 <GregNoel> You'd have to keep track of whether it was both a build dir and a src dir 18:23:11 <garyo-home> 1210: I don't think it's important enough to assign a resource now. (And Greg: yes, there are complications. I'm not voluteering...) 18:23:19 <stevenknight> right, and since so much dir behavior is implicit, 18:23:39 <stevenknight> figuring out what is or isn't "build dir" vs. "src dir" has the potential for odd heuristics 18:23:50 <GregNoel> Since Steven is about to turn into a pumpkin, this is a bad time to be designing. 18:23:55 <stevenknight> 1210: 2.x, future draft pick 18:24:01 <GregNoel> done 18:24:43 <stevenknight> 1213: research (decide if it's still a problem) 18:24:50 <GregNoel> 1213, bypass for today? 18:24:56 <garyo-home> Yes, I'll research that one (1213) 18:24:57 <stevenknight> sure, bypass 18:25:03 <stevenknight> er, i meant, garyo! 18:25:04 <GregNoel> which? 18:25:09 <stevenknight> 1213 18:25:15 <stevenknight> 1213 garyo, done 18:25:16 <garyo-home> I'll look at it. 18:25:18 <garyo-home> quick test. 18:25:26 <GregNoel> ok, garyo, research 18:25:33 <stevenknight> 8: greg, you want it to see the -include in CCFLAGS and deduce the dependency? 18:25:59 <GregNoel> yes 18:25:56 <stevenknight> there's another issue i just looked at that takes the other approach, 18:26:04 <stevenknight> adding a FORCEINCLUDE variable 18:26:12 <stevenknight> that you set and then it adds the -include flag to the command line 18:26:19 <GregNoel> hmmm.... 18:26:21 <stevenknight> like we do with CPPPATH and -I options e.g. 18:26:26 <stevenknight> i think that's more "sconsy" 18:26:37 <garyo-home> Agree w/ steven 18:26:50 <GregNoel> Would have to be parsed by MergeFlags()... 18:27:11 <garyo-home> I don't like looking at cmd lines to figure out deps 18:26:57 <stevenknight> okay, fifteen seconds... 18:27:02 <stevenknight> agreed re: MergeFlags() 18:27:08 <stevenknight> two weeks from now, I assume 18:27:14 <stevenknight> for next meeting 18:27:15 <garyo-home> ok see you then! 18:27:16 <stevenknight> later guys... 18:27:18 <GregNoel> bye 18:27:20 * stevenknight has quit ("Leaving") 18:27:21 <garyo-home> bye 18:27:26 <garyo-home> well that was short. 18:27:40 <GregNoel> Yeah, but we cleared the current issues 18:27:50 <GregNoel> I guess the meeting time is still a work in progress 18:27:58 <garyo-home> Greg, would you like to help me edit the 2006h2 sshet? Send me a direct invite or something? 18:28:16 <garyo-home> Yes, time is a little early for me but we're hoping to get some Euro zone folks maybe? 18:28:27 <GregNoel> ok, I'll do that. I'd really like to figure out why it's failing. 18:28:44 <garyo-home> The current issue sheet I never have any trouble with. 18:29:00 <garyo-home> But 2006h2 has never worked for me ever. 18:29:02 <GregNoel> I know there's a long timeout if you view it and just close the page 18:29:33 <garyo-home> Could be it, but I don't think so. 18:29:42 <GregNoel> without going through the file->close selection, I mean 18:30:33 <garyo-home> Yes, I know. (Just to be clear I'm talking about 2005*h2 even though I said '06 above) 18:31:13 <garyo-home> Maybe a direct invite will work. 18:31:28 <GregNoel> I'll send you an individual invitation and we'll see if that works. (I wonder if invitations age? That's from six months ago...) 18:31:39 <garyo-home> Could be. 18:32:00 <GregNoel> OK, I'll do that 18:32:09 <garyo-home> OK, well if we're done for now maybe I'll try to make some progress on a bug or two. And I'll add that doc line you were asking for :-) 18:33:05 <GregNoel> Yes... Strangely enough, I wanted a rule like he described, and was annoyed that there was no way to do it. This was a most elegant solution and should be documented. 18:33:57 <garyo-home> Will do. See you on the ML... 18:34:06 <GregNoel> G'night 18:34:10 <garyo-home> bye now 18:34:15 * GregNoel has been marked as being away 19:54:41 * garyo-home has quit ("ChatZilla 0.9.83 [Firefox 3.0.3/2008092417]")