Wiki

Clone wiki

SCons / BugParty / IrcLog2008-07-21

17:18:02 * bdbaddog (n=bdeegan@adsl-71-131-30-2.dsl.sntc01.pacbell.net) has joined #scons 18:51:06 * stevenknight (n=stevenkn@69.36.227.131) has joined #scons 18:51:35 * GregoryNoel is no longer marked as being away 18:51:48 <GregoryNoel> Hey, Steven... 18:53:59 <stevenknight> hi greg 18:54:26 <GregoryNoel> I hope Gary can make it; he sounded dragged out in his note. 18:54:30 <stevenknight> on the late shuttle tonight, so i turn into a pumpkin early :-( 18:54:36 <stevenknight> agreed re: Gary 18:55:22 <stevenknight> he said he'd updated the spreadsheet, but i only see his comments on early issues 18:56:05 * GregoryNoel is still pulling up the spreadsheet 18:59:22 <GregoryNoel> Ah, Brandon is adding comments. But you're right; I don't see Gary's comments past the initial few. 19:00:34 <GregoryNoel> Brandon, are you here for the bug meeting? 19:00:46 <GregoryNoel> Or anyone else? 19:02:48 <Azverkan> I'm here 19:02:52 <Azverkan> was still typing 19:03:00 <GregoryNoel> Hey, Brandon... 19:03:10 <stevenknight> hi brandon 19:03:43 <Azverkan> hey 19:04:04 <GregoryNoel> Can you keep going, but split your attention here? 19:04:09 <Azverkan> yep 19:04:19 <GregoryNoel> OK, shall we proceed? 19:04:23 <stevenknight> since we have three + gary's comments, and the clock's ticking, shall we get started then? 19:04:26 <stevenknight> yes 19:04:35 <stevenknight> 2133: 19:05:01 <stevenknight> the more that i think about it, i'm on the fence like gary 19:05:44 <GregoryNoel> I think I'd move from INVALID to WONTFIX, but I still think it's not a bug 19:05:44 <stevenknight> but it's still pretty common and not unreasonable usage 19:06:00 <GregoryNoel> Really? What's the use case? 19:06:16 <stevenknight> hmm, well maybe not common per se 19:06:21 <stevenknight> just surprising 19:06:38 <stevenknight> you want to run a post-processing script after building the target 19:06:48 <stevenknight> the post-processing script is built locally in your tree 19:07:09 <stevenknight> you AddPostAction() the script and expect that SCons will make sure it's built before it tries to execute it 19:07:16 <stevenknight> like it does with the regular Actions 19:07:39 <GregoryNoel> Yeah, but the next run, the post-processing doesn't get done since the command is already built. 19:07:36 <stevenknight> i guess it comes down to whether or not people should think an Action is an Action is an Action 19:07:54 <stevenknight> regardless of whether a Builder or AddPostAction associates it with the target 19:08:21 <GregoryNoel> I think it's bad design, for sure 19:08:32 <stevenknight> AddPostAction() in general? I agree 19:08:44 <GregoryNoel> No, this use case, running a command as part of its own build. 19:08:50 <stevenknight> ah 19:09:04 <GregoryNoel> I just don't see any real reason to run the command just after building it and no other time. 19:09:05 <Azverkan> Feels more like a side effect of how builders are implemented that being a required feature to me 19:09:47 <stevenknight> I don't know, i do see people using it for things like unit test executables 19:10:03 <Azverkan> You have two types of these post actions, some that you want to always run and some that you only want to run after the executable changes 19:10:09 <GregoryNoel> For unit tests, you can use a synthetic target 19:10:12 <stevenknight> but you can argue that's just because we do a lousy job with tests right now 19:10:27 <Azverkan> and in both cases they may or may not have real targets, but fake targets like the windows registry getting updated etc... 19:10:37 * stevenknight agrees with Azverkan 19:10:47 <Azverkan> I think the real fix is to make Alias() more interchangeable with File() 19:11:03 <GregoryNoel> yes 19:11:05 <stevenknight> yes yes yes 19:11:07 <Azverkan> there are some cases where you have to make a File() that does not exist to work around limitations in Alias() 19:11:21 <GregoryNoel> yes 19:11:41 <stevenknight> okay, i can see WONTFIXing this in favor of a more comprehensive solution involving being able to use Alias 19:11:48 <stevenknight> as both a source and a real Dependency 19:11:53 <GregoryNoel> I'll go for that. 19:12:04 <stevenknight> like Brandon said 19:12:16 <stevenknight> okay, done 19:13:00 <stevenknight> I'll open a new one to track the Alias-wherever-you-can-use-File enhancement 19:12:51 <GregoryNoel> 2134 19:13:03 <stevenknight> 2134: 19:13:45 <stevenknight> Greg, did you have a particular 2.x feature in mind to make this easier? 19:13:46 <GregoryNoel> 2133 spinoff: "Synthetic Targets" 19:13:55 <stevenknight> oh, wait, just saw it in your comment 19:13:57 <GregoryNoel> 'attribute' 19:14:30 <stevenknight> gotta transfer buses, expect short disconnect 19:14:40 * stevenknight has quit ("Leaving") 19:16:35 * stevenknight (n=stevenkn@69.36.227.135) has joined #scons 19:16:41 <GregoryNoel> I just don't know if it's worth trying to implement in the 1.x timeframe if we're just going to reengineer it in 2.x, especially since 1.x is already getting so full. 19:16:47 <stevenknight> and we're back... 19:17:11 <stevenknight> yeah, 1.x is definitely full 19:17:19 <stevenknight> no one's beating down the doors for this 19:17:25 <stevenknight> let's go ahead and push it out to 2.x 19:17:30 <stevenknight> (assume we're still on 2134) 19:17:37 <GregoryNoel> works for me (yes) 19:17:57 <stevenknight> okay, 2134, 2.x, p3 (?) 19:18:10 <GregoryNoel> yes, issues@scons 19:18:25 <stevenknight> 2135: consensus 1.0.x p2 19:18:34 <GregoryNoel> done 19:18:41 <stevenknight> 2136: consensus 1.x p2 19:18:46 <GregoryNoel> done 19:19:00 <stevenknight> 2137: consensus 1.0 p3 19:19:08 <GregoryNoel> yes, but who? 19:19:49 <stevenknight> guess it partly depends on how soon we push out 1.0 19:20:12 <GregoryNoel> I think it should be within a week; it's aged enough. 19:20:12 <stevenknight> i'll take it, unless you have any text you've already started 19:20:22 <stevenknight> agree re: aged enough 19:20:31 <GregoryNoel> no, in fact, I think I'm too close to it 19:20:27 <stevenknight> 2137: 1.0 p3 stevenknight 19:20:38 <GregoryNoel> done 19:21:01 <GregoryNoel> 2138 19:21:12 <stevenknight> 2138: consensus 1.0.x p2 19:21:13 <stevenknight> me 19:21:22 <GregoryNoel> ok, works 19:21:39 <stevenknight> 2140: changed my mind, 2.x p4 19:21:50 <stevenknight> move it up if someone actually comes up with a good interface for it 19:21:58 <stevenknight> (the underlying hook, i mean) 19:22:10 <GregoryNoel> okay 19:22:37 <stevenknight> 2141: 1.0.x p2, me 19:22:40 <GregoryNoel> 2141, consensus 19:22:47 <Azverkan> 2140: I'd think that the hook would probably be driven more by distributed build requirements like IncrediBuild than the actual submitted bug 19:23:18 <stevenknight> 2140: agree that that's a more compelling reason than this particular use case 19:23:37 <stevenknight> but someone still has to care enough to pony up the code 19:24:05 * Azverkan is at the moment I'm worried about the GIL not getting fixed in Py3k 19:24:10 <stevenknight> 2142: consensus dup 19:24:35 <GregoryNoel> 2142, I added a comment to 2132 asking that he takes makes sure it uses env[ENV] 19:24:29 <stevenknight> 2143: consensus 1.x p2 david 19:24:46 <GregoryNoel> 2143, done 19:24:47 <stevenknight> cool, thanks 19:24:52 <stevenknight> (re: 2142) 19:25:36 <stevenknight> 2144: Brandon, can you say more about the Windows API issue at work here 19:26:01 <Azverkan> It's probably more complicated than we want to spend on it 19:26:01 <GregoryNoel> 2144, I agree with Steven's comment, but we should ask Benoit about it. 19:26:16 <Azverkan> But basically Ctrl-C event and job trees is the root of the issue 19:26:31 <stevenknight> on the discussion or on actually fixing it (or working around it) in SCons? 19:26:32 <Azverkan> Unix sends signals to subprocesses differently than windows does 19:26:39 <GregoryNoel> Is the original issue from Windoze? 19:26:46 <stevenknight> yes, IIRC 19:26:58 <Azverkan> yeah 19:27:16 <Azverkan> it gets even nastier if you have a scons running inside a scons running inside a scons or something like that 19:27:16 <GregoryNoel> Then it looks like Brandon is volunteering.... {;-} 19:27:33 <Azverkan> I dont think it's something we can ever fix on the command line without an IPC layer 19:28:11 <Azverkan> but the named event hack I did in the past worked well enough that I could submit a patch 19:28:05 <stevenknight> basically something else wraps and just handles the interrupt 19:28:16 <stevenknight> and passes word in a controlled way to the back-end SCons process? 19:28:50 <Azverkan> in my case we just modified the gui to send the named event instead of the Ctrl-C or the Ctrl-Break which both have bugs (and different kinds) 19:29:04 <Azverkan> Ctrl-C corrupts scons and Ctrl-Break corrupts subtools 19:29:16 <GregoryNoel> Ah, Windo$e... 19:29:39 <stevenknight> and i assume the wrapper that handles the Ctrl-{C,Break} needs to be a separate task, not just a thread? 19:29:50 <Azverkan> yes and scons cannot be a child of that task 19:30:00 <Azverkan> so you have to spawn a task parented by the parent of scons 19:30:12 <stevenknight> what fun! 19:30:19 <Azverkan> detach the console from scons and reattach the console to that 19:30:27 <Azverkan> its probably not really worth the effort 19:30:32 * GregoryNoel wonders about Steven's idea of fun 19:30:48 <Azverkan> I'd put it off until we think about releasing 2.x 19:31:02 <GregoryNoel> Future? Or wontfix? 19:31:11 <stevenknight> future 19:31:17 <Azverkan> future because you will get new bugs otherwise 19:31:21 <Azverkan> over and over 19:31:31 <stevenknight> Brandon, could you add a write up describing the above so it gets captured? 19:31:32 <GregoryNoel> ok, what priority then? 19:31:39 <Azverkan> yeah 19:31:44 <stevenknight> thanks 19:31:46 <stevenknight> i'd say p2 19:31:56 <stevenknight> agree it sounds like a lot of effort 19:32:06 <stevenknight> but it goes to the heart of the "correct build" priority 19:32:07 <GregoryNoel> ... I was thinking p3 but I'll go with p2 19:32:17 <GregoryNoel> ah, good point 19:32:25 <stevenknight> okay, 2144, future, p2 19:32:29 <GregoryNoel> done 19:32:48 <stevenknight> 2145: research, me 19:33:13 <GregoryNoel> ok, done 19:33:29 <stevenknight> 2146: anytime 19:33:36 <stevenknight> me if no one else volunteers 19:33:44 <GregoryNoel> done 19:33:59 <GregoryNoel> I think you're the only one who understands it 19:33:52 <stevenknight> re: 2146 19:34:45 <stevenknight> if we scrap the surrogate stuff, what about generatiing the output with scripts modeled after the test/.py infrastructure? 19:35:12 <GregoryNoel> say more? 19:35:52 <stevenknight> all of the in-line <scons_example> things 19:36:18 <stevenknight> get turned into a separate self-contained script based on the same API that the test/.py tests use 19:36:41 <stevenknight> generating the example output becomes a matter of running some script like runtest.py 19:36:51 <stevenknight> that captures the output at the right step 19:37:32 <GregoryNoel> I'm not sure... It still seems awkward. Maybe you should start a thread on the dev list. 19:37:43 <stevenknight> your ideas about making it a more integrated part of the packaging build are on target regardless of this internal implementation detail 19:37:56 <stevenknight> okay, i can do that 19:38:10 <GregoryNoel> ok, I'll look for your message 19:38:22 <stevenknight> shall we try to make some progress on 2006H2? 19:38:29 <GregoryNoel> How about that? We finished the current issues! 19:38:39 * stevenknight rejoices 19:38:36 <GregoryNoel> Yes, onward 19:38:46 <GregoryNoel> only a few of those left now... 19:39:13 <stevenknight> cool 19:39:23 <stevenknight> net lag pulling up the spreadsheet... 19:39:25 <stevenknight> there we go 19:39:35 <bdbaddog> :) 19:39:36 <stevenknight> where did we leave off? 19:39:42 <GregoryNoel> 1437 is first 19:39:55 <stevenknight> 1437: consensus dup 19:39:57 <stevenknight> hey bill 19:40:23 <bdbaddog> Good evening. Gotta run in a few,but here for a few. 19:40:36 <stevenknight> cool, thanks 19:40:39 <GregoryNoel> Hey, Bill 19:40:42 <bdbaddog> timely 1437 is related to email thread Greg and I have been sharing. 19:40:49 <stevenknight> on 2006H2 19:40:49 <stevenknight> 1438: i 19:40:50 <GregoryNoel> yep 19:40:55 <stevenknight> try again 19:41:10 <stevenknight> 1438: I'm coming around to Bill's suggestion of 2.x and redoing this 19:42:09 <stevenknight> I'm starting to think I can kill off VariantDir by making the repository support more flexible 19:42:20 <stevenknight> let you really stack directories arbitrarily 19:42:33 <GregoryNoel> I tried to go there once; you shot me down. 19:42:45 <stevenknight> yep, i was wrong 19:42:51 <bdbaddog> sounds like it would support very complicated schemes, but would it be easy to to the simple case? 19:43:33 <stevenknight> i think so 19:43:49 <bdbaddog> worth a wiki page to discuss ? 19:43:54 <bdbaddog> and/or email thread. 19:43:57 <GregoryNoel> I'd certainly like to see a proposal 19:44:00 <stevenknight> sounds like it 19:44:16 <Azverkan> I'm also of the opinion that the current approach needs to replaced and since that would potentially require scripts to be updated it would have to be a 2.x feature 19:44:17 <stevenknight> probably research, p3, me then 19:44:30 <bdbaddog> ok guys. gotta run. sorry to be a short timer tonight. 19:44:37 <GregoryNoel> your research? That's 'anytime' 19:44:48 <stevenknight> :-) 19:45:21 <GregoryNoel> I'm not kidding; I plan to manipulate the ordering so 'anytime' always sorts just after the current release. 19:45:35 <GregoryNoel> Er, the next immediate release. 19:45:47 <stevenknight> right, but i'm actually somewhat serious too 19:46:02 <stevenknight> i've adopted your terminology and agree that "research" should be higher priority 19:46:10 <stevenknight> for reclassification 19:46:18 <stevenknight> even if i'm not consistent about doing that 19:46:37 <GregoryNoel> Hmmm, ok, then which do you mean? 19:46:41 <stevenknight> so i'd rather see this as "research" so the proposal has at least a shot at getting written up sooner rather than later 19:46:53 <GregoryNoel> I'll go for that 19:46:56 <stevenknight> if only because I have it staring at me reminding me i haven't done it yet... 19:47:08 <GregoryNoel> {;-} 19:47:18 <stevenknight> okay, 1438: research, sk 19:47:22 <GregoryNoel> done 19:47:42 <GregoryNoel> 1439, toolchain 19:47:49 <stevenknight> done 19:47:55 <stevenknight> (I could go with you on invalid, too) 19:48:19 <stevenknight> 1442: toolchain / dup 1437 19:48:23 <stevenknight> ? 19:48:36 <GregoryNoel> yes, invalid, I was thinking of another issue 19:49:19 <stevenknight> okay, just to be clear: 1439: invalid ? 19:49:23 <Azverkan> not convinced that 1439 is invalid 19:49:32 <GregoryNoel> 1439, yes 19:49:38 <Azverkan> but a decision that the project sure either make for or against 19:50:00 <Azverkan> seems like the current approach is that it is not supported 19:50:52 <GregoryNoel> uh, wait, 1439 has been already taken care of 19:51:11 <Azverkan> yeah there is like 10 in a row I think 19:51:14 <stevenknight> whoa, hang on, i already have a comment on here from July 1 saying this was getting closed 19:51:46 <stevenknight> but it looks like I didn't change the status 19:52:06 <GregoryNoel> Yes, it's marked invalid; it's still just in the spreadsheet. 19:52:06 <stevenknight> yeah, we've been through a bunch of these already 19:52:23 <stevenknight> ah, we only have 7 left from 2006H2 19:52:25 <GregoryNoel> next is 1490 19:52:48 <stevenknight> damn, i'm down to the last minute 19:52:55 <stevenknight> same time next week? 19:53:00 <stevenknight> we obviously won't have gary 19:53:15 <stevenknight> brandon, is this time good for you -- it's really helpful to have you here 19:53:48 <Azverkan> the earliest I get home by is 7:00PM west coast 19:54:24 <Azverkan> the time we have now is best for me so far 19:54:34 <stevenknight> okay, sounds good 19:54:35 <GregoryNoel> You can certainly join a little late, but would some other time be better? 19:55:02 <Azverkan> my work hours are more or less demand driven 19:55:07 <stevenknight> if so, say the word 19:55:13 <Azverkan> not really 19:55:20 <stevenknight> i'll assume same time (19h00 PDT) next week unless i hear otherwise 19:55:27 <Azverkan> I'm other completely busy or free by 1900 19:55:35 <GregoryNoel> OK, then we demand that you're here next time {;-} 19:55:30 <stevenknight> ....and i'm at my stop 19:55:32 <stevenknight> later... 19:55:35 * stevenknight has quit ("Leaving") 19:56:19 <GregoryNoel> I've got to go, too; le Tour de France calls... 19:56:25 <GregoryNoel> cul 19:56:34 * GregoryNoel has been marked as being away 21:55:33 * bdbaddog (n=bdeegan@adsl-71-131-30-2.dsl.sntc01.pacbell.net) has left #scons

Updated