Clone wiki

SCons / BugParty / IrcLog2008-05-20

09:43:15 * bdbaddog ( has joined #scons 16:59:51 * stevenknight (n=stevenkn@nat/google/x-20f1d53866f4ad08) has joined #scons 16:59:53 <GregoryNoel> Who's here for the bug party? 16:59:59 <stevenknight> me me me 17:00:19 <bdbaddog> I though it was yesterday. 17:00:30 <stevenknight> we had some follow-up to do 17:00:38 <GregoryNoel> followup with the rest of the current issues tonight 17:01:23 <GregoryNoel> Gary is marked away; are you really here? 17:01:38 <stevenknight> you == ? 17:01:47 <GregoryNoel> you == Gary 17:02:14 <GregoryNoel> he said he'd be here; shall we wait a couple of minutes? 17:02:44 <stevenknight> sure, i can't see waiting as a problem 17:03:00 <stevenknight> we should have plenty of time 17:02:56 <GregoryNoel> Are you on the bus yet? 17:03:44 <stevenknight> doesn't come until 5:25 17:04:03 <stevenknight> i should only have a slight hiccough when i change from land-based wifi to mobile 17:04:28 <GregoryNoel> Bill, we'll pick up with 2047 in the current issues, if you can join us 17:04:40 <bdbaddog> sure. lemme clock out. 17:04:52 <garyo-home> Hi, I'm here now. 17:05:14 <bdbaddog> which spreadsheet are we looking at? 17:05:26 <GregoryNoel> current issues of current issues 17:06:35 <GregoryNoel> I'm going to grab a Coke while we're getting set up 17:06:38 <garyo-home> For 2047 we were discussing whether it could just be a warning, iirc 17:06:53 <garyo-home> at least for 1.0? 17:07:22 <stevenknight> bdbaddog: "Current issues" 17:07:44 <bdbaddog> got it, row 339 in the spreadshhet 17:07:49 <garyo-home> Hey, where did the queries in the BugParty page go? 17:08:00 <GregoryNoel> ??? 17:08:09 <GregoryNoel> Still there, as far as I know. 17:08:15 <stevenknight> re: 2047, we were just zeroing in on me taking a look at how impactive it would be to turn them back into warnings 17:08:32 <stevenknight> or not back, since this is actually a new error message 17:08:51 <garyo-home> oh yeah, duh 17:09:23 <garyo-home> Steven: any sense of what would happen if we just warn and let it try to link? 17:09:24 <bdbaddog> seems like a regression, sounds like making it a warning would be the right thing to do. if it can work in some toolchains. 17:09:50 <garyo-home> (I guess it has to choose C++ or Fortran; that's the hard part) 17:09:51 <stevenknight> exactly 17:10:02 <stevenknight> they may get an error if the toolchain doesn't allow this interoperability 17:10:08 <stevenknight> which is what David was trying to guard against 17:10:23 <stevenknight> but a little drastically 17:10:26 <GregoryNoel> In the long run, it should be made 'smarter still' but is that this bug or another one? 17:10:26 <garyo-home> Do you know which linker it used to use in that case? 17:10:33 <garyo-home> Another one, Greg. 17:10:41 <garyo-home> IMHO 17:10:54 <GregoryNoel> OK, I'll add it when I clear these out 17:10:54 <stevenknight> okay, give it me to make it a warning for 1.0 17:11:05 <stevenknight> and i'll delay it to later if it looks too impactive 17:10:59 <GregoryNoel> done 17:11:09 <garyo-home> good. 17:11:56 <garyo-home> 2050 is hard I think. 17:12:08 <GregoryNoel> and nasty 17:12:11 <stevenknight> yeah 17:12:20 <stevenknight> just trying to get caught up with the code in it 17:12:46 <GregoryNoel> There's clearly a deadlock, but it's not clear how to break it for the child. 17:12:54 <garyo-home> What happens if you try 'from errno import ENOENT, ENOTDIR' at top level, so nothing happens in execvpe? 17:13:29 <GregoryNoel> In brief, the lock is created in the parent, so it's held by the child, leading to a deadlock, since the child will never release it. 17:13:55 <stevenknight> give this one to me as well, obviously 17:14:09 <garyo-home> The deadlock is in 'import' iiuc so maybe the fix is easy. 17:14:25 <GregoryNoel> Some of the problem is in the Python libraries, since they don't expect to fork at that point 17:14:29 <stevenknight> I can see if I can get Guido or Alex Martelli to advise 17:14:34 <bdbaddog> I think that could go into release notes? (2050) with a if you do this it my hurt warning, to be addressed later? 17:15:12 <garyo-home> Is there a reproducible testcase for this? 17:15:03 <stevenknight> one nagging thing is bothering me about Benoit's analysis here... 17:15:16 <stevenknight> it suggests that "import" itself isn't thread safe 17:15:28 <stevenknight> or "import" of certain very well-used modules 17:15:41 <stevenknight> if that were the case, I don't think SCons would be the only thing with these symptoms 17:15:38 <GregoryNoel> No, it's not fork-safe, not the same thing 17:15:53 <stevenknight> ah, right 17:16:00 <stevenknight> but still 17:16:11 <stevenknight> it's not like we're the only Python-based application that forks things 17:16:17 <bdbaddog> I thought that you couldn't run python builders in parallel because of the GIL 17:16:29 <bdbaddog> (aka in separate threads) 17:16:41 <bdbaddog> well. python threads. 17:16:55 <GregoryNoel> separate threads are in the same process so the lock will be released 17:17:31 <GregoryNoel> it's the child thread that thinks it has the lock but doesn't 17:18:01 <garyo-home> In any case, my opinion is it should be investigated more, but maybe not fixed til 1.x, p1. 17:18:10 <stevenknight> the thing that really concerns me is that in some of those cases (import within function) 17:18:13 <GregoryNoel> If Guido or Alex can help, that would be the way to go 17:18:31 <stevenknight> the import was moved thre because scoping rules weren't letting it get imported at the global module level 17:19:00 <stevenknight> i agree with gary: 1.x p1 17:19:07 <stevenknight> and with Greg re: getting help 17:19:05 <GregoryNoel> done 17:19:43 <GregoryNoel> 2051, consensus? 17:19:54 <garyo-home> 2051: consensus 1.x p3? 17:20:05 <stevenknight> i was looking at this again this morning 17:20:24 <stevenknight> since we don't supply an egg, he must have packaged/installed it himself 17:21:04 <stevenknight> it shouldn't hurt anything to add the code he's suggesting 17:21:29 <stevenknight> but the fact that "easy" install makes you do this sucks 17:21:46 <stevenknight> of course, we are doing something different than most Python packages 17:21:54 <stevenknight> okay, i'll stop ranting 17:22:09 <stevenknight> 1.x p3 17:22:23 <bdbaddog> 1.x p3 17:22:37 <garyo-home> sure 17:22:39 <GregoryNoel> done 17:22:44 <GregoryNoel> 2052 17:22:58 <stevenknight> gotta get on the bus, may have to reconnect 17:23:00 <garyo-home> 2052 must be trivial? 17:23:01 * stevenknight has quit ("This computer has gone to sleep") 17:23:38 <GregoryNoel> I don't use M$'s Java wanabe, so I have no clue 17:23:51 <garyo-home> Probably just forgetting to check the length of the parsed version strings or something. 17:24:02 <garyo-home> GregoryNoel: :-) 17:24:32 <garyo-home> I bet I could fix that, give it to me. I think I have a machine with one of those .NET versions on it. 17:24:52 <GregoryNoel> works for me 17:24:57 <bdbaddog> +1 17:25:01 <GregoryNoel> 1.x? 17:25:21 * sgk_ (n=stevenkn@ has joined #scons 17:25:31 <sgk_> okay, i'm back 17:25:40 <sgk_> what are we up to? 17:25:43 <garyo-home> I'd say 1.x, but the short-named .NET versions are the real releases... 17:25:44 <bdbaddog> 2052 17:25:57 <GregoryNoel> Gary wants 2052, but not settled on milestone or priority 17:26:00 <bdbaddog> .net version number issue. 17:26:01 <garyo-home> so it may bite us if not fixed in 1.0. 17:26:13 <sgk_> right 17:26:16 <bdbaddog> yeah. if it's trivial 1.0 17:26:30 <sgk_> gary, do you want to take a look for 1.0 and back off if it looks risky? 17:26:39 <garyo-home> ok, sorry I have to go guys, we're almost done with that spreadsheet anyway. 17:26:54 <sgk_> okay, thanks 17:27:19 <GregoryNoel> I'll set it to 1.x p1 17:27:27 <sgk_> 1.0 p2, gary to back off if it's risky 17:27:38 <sgk_> okay, p1 17:28:01 <GregoryNoel> OK, I guess. 17:28:05 <GregoryNoel> 2056? 17:28:04 <sgk_> BTW, i have a question re: our target milestones 17:28:17 <sgk_> did we change our interpretation? 17:28:26 <GregoryNoel> Not as far as I know 17:28:34 <sgk_> yesterday and today we are treating them as "1.0 means it should go in for 1.0" 17:28:40 <sgk_> right? 17:28:43 <GregoryNoel> Yes 17:29:11 <GregoryNoel> 1.x (note the x) means "during the 1.0 cycles" 17:29:12 <sgk_> but we've got a whole bunch of 1.0 target milestone things already that clearly aren't happening in the slim time between 0.98.x and 1.0 17:29:44 <GregoryNoel> I didn't put them there; others decided they should be; I've pushed for 1.x 17:30:13 <GregoryNoel> but I basically agree with your point 17:30:18 <sgk_> oh, wait, i didn't look closely enough 17:30:18 <bdbaddog> why don't we wrap up the spreadsheet and take a quick look at what's left for 1.0 ? 17:30:38 <GregoryNoel> I'd suggest that the next bug party triage 1.0 more closely. 17:30:40 <sgk_> most of these are doc issues which we did agree we'd work on while 0.98.x is soaking 17:30:55 <sgk_> yeah, sorry to derail 17:30:59 <sgk_> back to 2056 17:31:20 <bdbaddog> I know there was traffic on the mailing list for that, did someone propose a fix? 17:31:44 <GregoryNoel> I'm not opposed to a 0.98.5, but six release candidates is a lot 17:31:53 <sgk_> i found a guy's blog with a technique for propagating %ERRORLEVEL% even when using setlocal+endlocal 17:32:28 <bdbaddog> 6 RC's better than 1.0, 1.0.1, 1.0.2 in a week though.. :) 17:32:35 <sgk_> and we don't have a test case that explicitly tests scons.bat for things like this 17:32:43 <GregoryNoel> bdbaddog: point taken 17:32:46 <sgk_> just the scons wrapper itself 17:33:14 <sgk_> me, 1.0 (or 0.98.5), p1 17:33:23 <bdbaddog> 1.0p1 sounds good to me. 17:33:26 <GregoryNoel> done 17:34:02 <sgk_> 2057: the same big issue it's always been 17:34:12 <sgk_> looks like consensus is 1.x and p3 17:34:24 <GregoryNoel> yeah, but who? 17:34:41 <GregoryNoel> someone needs to think about the criteria and write them up 17:34:57 <sgk_> i'll take it 17:35:13 <GregoryNoel> no no no, someone else; you know too much about the internals 17:35:21 <sgk_> ok... 17:35:27 * sgk_ is properly chastened... :-) 17:35:43 * GregoryNoel couldn't spell that on a bet 17:36:15 <bdbaddog> steven how ugly will it be to understand the code issueing that message? 17:36:29 <GregoryNoel> Nasty 17:36:47 <bdbaddog> nasty understand everything, or understand 1/2 of everything? 17:37:07 <sgk_> it does get a little involved 17:37:20 <sgk_> but what about writing up requirements not from a code perspective 17:37:24 <sgk_> but purely from a user perspective 17:37:24 <GregoryNoel> The underlying problem is what do do if a Builder isn't configured, so there's no way to recognize the suffix 17:37:49 <sgk_> then refine that 17:38:03 <bdbaddog> so if I have file and nobodies registered .xyz, what to do? 17:38:07 <sgk_> that might be fruitful without having to grok a whole bunch of the code 17:38:15 <GregoryNoel> bdbaddog: exactly 17:38:32 <GregoryNoel> ("nobody's") 17:38:59 <bdbaddog> GregoryNoel: you are correct sir. :) 17:39:17 <GregoryNoel> I guess I can draft something, but I'm not sure I'd get all the nuances 17:39:57 <GregoryNoel> bdbaddog: don't call me "sir" -- I worked for a living! {;-} 17:39:57 <sgk_> that's okay, if it gets us closer to the goal 17:40:20 <GregoryNoel> OK, me, when? 1.x? 17:40:26 <bdbaddog> Greg I'd be willing to be sounding board on that for you. 17:40:40 <GregoryNoel> Thanks 17:40:45 <bdbaddog> 1.x 17:40:58 <GregoryNoel> although I'm sure there will be lots of drafts :-( 17:41:07 <bdbaddog> google docs is your friend.. :) 17:41:23 <GregoryNoel> or the wiki... 17:41:49 <bdbaddog> :) 17:41:54 <GregoryNoel> onward? 17:41:56 <sgk_> yes 17:42:12 <sgk_> that's it for the spreadsheet 17:42:29 <sgk_> but i think a few more have come in since it was generated 17:42:31 <GregoryNoel> last one, 2058? 17:42:32 <sgk_> hang on... 17:42:37 <sgk_> oh, right 17:43:05 <GregoryNoel> I haven't looked at the new logos he developed, but the first set looked good 17:43:15 <sgk_> cool 17:43:36 <sgk_> i'd say let him check in directly 17:43:40 <GregoryNoel> I think it should go to Gary to set up a web page for them 17:44:01 <sgk_> cool, that sounds better 17:43:59 <GregoryNoel> but the problem is setting a policy and a license for them 17:44:18 <sgk_> aha 17:44:34 <bdbaddog> do we need to file a trademark on it? 17:44:50 <GregoryNoel> possibly 17:45:02 <sgk_> not sure 17:45:01 <GregoryNoel> need legal advice, for sure 17:45:15 <sgk_> we can check with the SFLC, we're "clients" 17:45:28 <GregoryNoel> yes, but you have to do that 17:45:55 <sgk_> okay, then make this one a task for Gary to set up the web site 17:46:13 <sgk_> and how about another task for me to check with SFLC re: trademark and/or licensing issues 17:46:29 <GregoryNoel> done 17:46:11 <GregoryNoel> And another related issue, probably needs to be well-started by 1.0, is getting releases on file for everyone who has ever submitted code 17:47:13 <GregoryNoel> and I'll create a task for the disclaimers as well 17:47:10 <sgk_> re: releases: we're helped by our non-restrictive license 17:47:29 <sgk_> contributors don't have to actually assign code ownership to us 17:47:53 <sgk_> they just have to license it to us 17:48:08 <sgk_> that tends to make corporate lawyers less jumpy 17:48:22 <GregoryNoel> and exactly who has done that so far? 17:48:29 <GregoryNoel> anybody? 17:48:35 <sgk_> a handful 17:48:47 <GregoryNoel> didn't know that 17:48:54 <sgk_> or rather, a handful have actually assigned ownership 17:48:54 <sgk_> early contributors like Anthony and Charles 17:49:15 <sgk_> I think Gary sent me one as well 17:49:22 <bdbaddog> Should we put up a page listing contributions and assignments? 17:49:37 <sgk_> not a bad idea 17:49:50 <GregoryNoel> hmmm... not so sure 17:50:08 <sgk_> ? 17:50:09 <bdbaddog> do all the svn comments indicate who contributed code as most was checked in by Steven? 17:51:12 <sgk_> probably upwards of 90%-95% (or more) owing to occasional lapses 17:52:01 <GregoryNoel> but there are hundreds of people; if we miss any, it could be a problem 17:52:13 <sgk_> legal or PR? 17:52:17 <bdbaddog> BTW I just check tigris, there are 39 1.0 bugs. most docs. 17:52:20 <bdbaddog> Legal.. 17:52:40 <bdbaddog> look at the driver issue between linux and openbsd. big stink. 17:53:22 <bdbaddog> Steven, would all the patches be in the mailing list archive? 17:53:28 <sgk_> yes, should be 17:53:44 <GregoryNoel> probably the starting point is to try to get a list of the contributors; the release notes should help 17:54:17 <sgk_> iirc, i think we've got 150+ unique names in the change log 17:54:20 <bdbaddog> Maybe float an email to users and dev mailing list asking for assignment or licensing from any patches contributed? 17:54:57 <GregoryNoel> Yes, for those who are still with us 17:55:08 <sgk_> right 17:55:14 <GregoryNoel> but I'll bet more than half will have to be tracked down 17:55:35 <sgk_> easily 17:56:13 <sgk_> but we're pretty penny-ante, and it helps to at least be able to show good faith efforts to try to get approval 17:55:57 <GregoryNoel> Mozilla had to go through this; do we have any contacts with them that far back? 17:56:27 <sgk_> with Mozilla? not directly 17:56:41 <sgk_> certainly not related to any SCons work 17:56:38 <GregoryNoel> They tracked me down through three address changes; not an easy job 17:56:46 <sgk_> wow 17:57:16 <GregoryNoel> good faith is one thing, but I hope we won't be penny-ante forever 17:57:26 <bdbaddog> :) 17:57:57 <sgk_> good point... 17:58:31 <sgk_> well, i can dig up the existing releases 17:58:41 <sgk_> and we can start a spreadsheet tracking everyone 17:58:53 <GregoryNoel> I know Mozilla went to the pain of removing any code they couldn't resolve 17:59:13 <GregoryNoel> and reverse-engineering any they wanted to keep 17:59:13 <bdbaddog> yes. and you have to do that clean room, it can be a huge pain. 17:59:15 <sgk_> the lawyer at SFLC said in practice if you get the big contributors you're pretty safe 17:59:48 <sgk_> i can make that discussion available too (email) 17:59:50 <GregoryNoel> I'm sure that would cover 99.44% of the code, but it only takes one 18:00:27 <sgk_> sure 18:00:29 <GregoryNoel> send a form to me; I haven't signed one yet 18:00:33 <bdbaddog> Well lets do what we can, and perhaps when Steven chats with SFLC about the icon, he can bring up the issue? 18:00:34 <sgk_> it can help w/prioritizing though 18:01:23 <GregoryNoel> how? 18:01:58 <sgk_> start with the biggest contributors 18:02:12 <GregoryNoel> yes 18:03:32 <GregoryNoel> Long silence; have we said all that can be said now? 18:03:46 <sgk_> i think so 18:04:36 <GregoryNoel> OK, then there's probably not a lot of use starting on the next spreadsheet tonight, so I'm going to go get dinner 18:03:58 <sgk_> hang on, doing one last check for newer issues 18:04:53 <GregoryNoel> We'll get those next week 18:05:01 <bdbaddog> ko. 18:05:15 <sgk_> even if they might require another 0.98.x? 18:05:33 <sgk_> fair enough 18:06:03 <sgk_> if anything looks really urgent we can convene by ML or IRC as necessary 18:06:09 <sgk_> thanks... 18:06:09 <GregoryNoel> yep 18:06:23 <GregoryNoel> OK, see you guys later 18:06:31 <sgk_> later 18:06:38 <bdbaddog> l8r. 18:06:43 * bdbaddog ( has left #scons