IrcLog2008 06 23

18:53:02 * stevenknight (n=[email protected]) has joined #scons 19:01:03 good evening, anyone else here for bugs? 19:01:08 <GregNoel> Warning: The Internet connection in this condo is in the middle of the living room, so rugrats are in play, Action Heros are on the TV, relatives and food are scattered all over, and multiple distractions are, er, distracting. 19:01:26 Hi guys, I'm here. 19:01:44 <GregNoel> Are you there, Bill? 19:02:01 Greg: welcome back! 19:02:26 <GregNoel> Well, I'm only sorta back, but I'll try to hang in there. 19:02:31 * garyo-home wonders if this is how you do the IRC italic thing 19:02:36 * garyo-home realizes it is. 19:03:10 <GregNoel> What italic thing? 19:03:28 i bet garyo's IRC client display /me messages in italics...? 19:03:34 yes. 19:03:41 <GregNoel> Ah. 19:04:01 Seems to be how one writes about oneself in the third person. 19:04:18 * GregNoel is not so sure 19:04:18 So, we're waiting for Bill? 19:04:31 i'd say let's start and he can join 19:04:34 * garyo-home thinks that's funny 19:04:39 ok 19:04:41 <GregNoel> ok 19:04:47 * stevenknight thinks we're all bozos on this bus 19:04:52 We're starting with 2098 then? 19:05:05 Or is it 2105? 19:05:19 <GregNoel> 2098 was only updated yesterday 19:05:32 <GregNoel> so I imagine we should start with 2105 19:05:30 ok, 2105 then. 19:05:38 2105 19:06:00 <GregNoel> I am swayed by Gary's argument, but not convinced 19:06:03 I"m ok with doc for 1.0, but I think it ought to be made to work someday. 19:06:23 <GregNoel> So maybe you should take it? 19:06:54 OK, give it to me, I'll doc it for now and reassign as 1.x or 2.0 afterward. 19:06:59 <GregNoel> done 19:07:03 done 19:07:31 2106: 1.0.x, p3, steven? 19:07:36 works for me 19:07:44 <GregNoel> done 19:07:53 2107? 19:08:28 2107: agree w/Greg that we need a comprehensive solution 19:08:31 I think Install as is should copy the source, but there should be a new way to do a real install. 19:08:41 define "real install?" 19:08:53 <GregNoel> I don't know how to get there from here 19:08:54 you mean like a package install? 19:09:00 real = like the BSD install command that takes mode/owner/group. 19:09:08 ah 19:09:20 why not do that with additional args to the current Install()? 19:09:26 <GregNoel> Configure makes the same distinction 19:09:42 Steven: I'd be OK w/ that. 19:10:07 I think it's already confusing enough to a lot of people that our Install() is modeled after the BSD command 19:10:17 Not sure if it's a + or - that those args could be set in the env though. 19:10:35 instead of something that means "make this part of the installation items for this package" like you have in a RPM or Deb linux distribution 19:11:04 Right; we should check Maciej's stuff, I'm pretty sure he solves this. 19:11:05 <GregNoel> Maybe we could split off InstallData/InstallExec? 19:11:21 good point re: Maciej 19:11:26 Greg: there should be a layer in between, but that's the right idea. 19:11:27 and the need for Install{Data,Exec} 19:11:49 <GregNoel> Maciej added the install prefix but that would kill anyone who used it for a copy 19:12:03 how about i take this one 19:12:14 <GregNoel> when? 19:12:18 OK w/ me. Research or ??? 19:12:23 i've been pretty sure integrating Maciej's stuff would fall to me anyway 19:12:42 1.x at the earliest, but no later than 2.x i'd think 19:12:44 <GregNoel> 1.x p4? 19:12:49 So 1.x p3/p4? 19:12:52 yeah, that sounds about right 19:12:57 1.x p4 19:12:59 <GregNoel> done 19:13:12 2108: trivial 19:13:17 <GregNoel> and consensus 19:13:20 yes 19:13:21 done 19:13:29 2109, what is im_func? 19:13:49 <GregNoel> dunno 19:13:57 Anyway looks like consensus is 1.0.x p2 Benoit? 19:14:00 it's an attribute on one of the Python data structures that refers the actual code function object 19:14:18 <GregNoel> Ah, yes, a 2.5 feature 19:14:33 trying to get the code function object (which has the compiled byte code) to calculate the signature is pretty involved 19:14:47 <GregNoel> so we need a backward-compatible shim? 19:14:59 It looks like a string action though, in the bug report. 19:15:04 you have to thread your way through different attributes depending on whether it's a function, or a callable object, and a couple of other non-obvious cases 19:15:22 <GregNoel> Benoit then 19:15:27 i think it just needs a little triage to isolate the difference in the reporter's Python version 19:15:29 ok w/ me 19:15:42 yeah 1.0.x p2 Benoit 19:15:47 <GregNoel> done 19:16:05 <GregNoel> 2110 19:16:11 2110: consensus 1.x p3 steven, unless tricky? 19:16:11 1.x p3 me 19:16:17 <GregNoel> done 19:16:17 yes 19:16:30 2111: dup of 2051 19:16:39 <GregNoel> ok 19:16:55 done 19:17:16 2112: consensus? 19:17:23 <GregNoel> ok, who? 19:17:21 2112: how have we survived this long with a summary line that violates the spec? 19:17:37 <GregNoel> Short entries? 19:17:42 must be 19:17:45 i'll take it 19:17:54 <GregNoel> ok, done 19:17:57 Steven, good question. Maybe someone's rpmbuild is less forgiving 19:18:38 2113: consensus 1.x p3? Could be earlier, it's likely to be easy 19:19:15 2113: how about 1.0.x p4 then? 19:19:23 fine w /me. 19:19:41 <GregNoel> I'm easy 19:19:55 OK, good progress! 19:19:58 ok, done 19:20:09 oh, wait -- who? 19:20:26 I could do it if you want. 19:20:34 works for me 19:20:52 ok. 19:20:59 on to 2007 q1? 19:21:14 I'm ready, looks like it starts w/ 1525. 19:21:35 ... which is clearly toolchain. 19:21:37 <GregNoel> still shuffling, not as easy as on my desktop 19:21:40 1525: consensus future+toolchain 19:21:45 devil's advocate, though: 19:22:08 it would actually be pretty trivial to just add some variables for these instead of hard-coding them in Platform/init.py 19:22:28 if the toolchain refactoring is going to take a while, is it worth doing something like that to help people in the meantime? 19:22:46 Yes, that's actually how I got interested in that. 19:22:48 or does that run the risk of boxing us into Yet Another feature that we'll have to maintain backwards compatibility for? 19:22:58 <GregNoel> hard choice 19:22:59 Started looking at how to expose those vars. 19:23:38 But you're right, given actual hours to be spent, toolchain refactor is going to take a while. 19:23:43 <GregNoel> I'd prefer to know where we're going before making short-term mods 19:24:17 Greg: that's true for sure. But how long will even the design part take? 19:24:42 I think once 1.0 is out we should spend some serious time on it. Not that I have any :-) 19:24:50 okay, so for this bug, let's leave it future+toolchain 19:24:51 <GregNoel> I've got a few updates at home, but design is always a long process 19:25:10 <GregNoel> OK, and if we can factor some out short-term, we'll do it. 19:25:19 with a notation to the effect that one early subtask in that should be nailing down the configurability interface 19:25:28 (i.e. variable names) 19:25:37 <GregNoel> (Actually, toolchain has been 2.x p4 I think.) 19:25:43 and retrofit that part to the existing code base if practical 19:25:43 Right, or maybe a simple functional interface, whatever. 19:25:52 yes. 19:25:53 <GregNoel> works for me 19:25:56 done 19:26:30 1538: fixed by Gary? 19:26:34 yes. 19:26:39 done 19:26:57 <GregNoel> Any patch for 1546? 19:27:01 1546, Ada? 19:27:47 No idea, let's mark it 2.x until other Ada requests come in or people vote for it. 19:27:54 ++ 19:28:00 <GregNoel> ++ 19:28:10 <GregNoel> p3? 19:28:19 Sure. 19:28:29 <GregNoel> done 19:28:38 1553: consensus=worksforme 19:28:43 done 19:28:58 1558: ??? 19:29:21 Would be cool, but does anyone understand pdb? 19:29:25 <GregNoel> Not me 19:29:34 just a smidge 19:29:50 only enough to put in place what we currently have 19:29:58 I think maybe 1.x or 2.0 p4. 19:30:02 ...and look at how well that's turned out... :-) 19:30:05 :-) 19:30:27 Better debugging in general would be nice 19:30:34 <GregNoel> agreed 19:30:42 how about p3? i'm swayed by your argument in the spreadsheet about making it easier to hack scons 19:31:09 Well, I'd use it if it were there, for sure. 19:31:20 So p3 is OK w/ me. 19:31:20 <GregNoel> 2.x p3, then? 19:31:43 1.x p3, i'd rather at least consider it sooner rather than later? 19:31:52 OK. If anyone with pdb knowledge turns up, we ask them to work on it. 19:31:55 <GregNoel> Hmmmm... 1.x p4. 19:32:04 i can go with that 19:32:06 ok compromise. 19:32:09 <GregNoel> done 19:32:11 done 19:32:32 <GregNoel> How's Nathan? 19:32:39 1567: awol, and i haven't followed up 19:32:58 1567: no sooner than 2.x unless Nathan is found. 19:33:10 <GregNoel> agreed 19:33:19 2.x, p...3? 19:33:25 <GregNoel> yes 19:33:28 And besides it'd have to be customized for each distro, yuck. 19:33:35 2.x p3 ok. 19:33:40 done 19:33:40 <GregNoel> but we need to mark it somehow so that 19:33:56 <GregNoel> we can find it again if Nathan shows 19:34:18 hmm, gsoc keyword? 19:34:31 <GregNoel> ok, would work 19:35:19 Not sure what we can do about 1570; no testcase. 19:35:29 i'm okay with closing it out 19:35:52 if it's important enough someone else will open up another issue with a testcase 19:35:54 agree. 19:36:02 <GregNoel> concur 19:36:23 1571: consensus 2.x p3 19:36:23 who? 19:36:40 or we don't need to assign 2.x -- i keep forgetting 19:36:48 Let's not. 19:36:53 <GregNoel> noone for now 19:37:02 done 19:37:13 1574: research, VisualStudio, me 19:37:18 <GregNoel> done 19:37:31 1575: 1.x p3 jim 19:37:43 ok 19:37:50 <GregNoel> done 19:38:23 1577: Greg, are you sure about .sconsign and test output? 19:38:26 * GregNoel has a rugrat on his head and other minor distractions.... 19:38:43 i hear rugrats are good eatin' 19:39:13 it's quiet here on the right coast 19:39:46 1577: the Configure stuff has to store the result somewhere 19:39:54 <GregNoel> Ok, where in the .sconsign is it kept? 19:39:58 when you re-run it will tell you things like "... yes (cached)" 19:40:26 i'm not sure off the top of my head, that code makes my head hurt when I look at it 19:40:28 I don't understand how it works, but I have tests that say .... "v1.03" (cached) 19:40:39 where that v1.03 was output from a config-compiled binary. 19:40:57 <GregNoel> It seems to cache positive/negative results, but it doesn't capture command output, for example 19:41:52 * GregNoel now has a naked rugrat crawling on his head.... 19:41:55 gah. check this out from SConf.py: 19:42:13 # Because we take responsibility here for writing out our 19:42:13 # own .sconsign info (see SConfBuildTask.execute(), above), 19:42:13 # we override the store_info() method with a null place-holder 19:42:13 # so we really control how it gets written. 19:42:13 # Because we take responsibility here for writing out our 19:42:14 # own .sconsign info (see SConfBuildTask.execute(), above), 19:42:16 # we override the store_info() method with a null place-holder 19:42:18 # so we really control how it gets written. 19:42:25 oops, sorry for the dup, didn't realize it was already in my buffer 19:42:38 <GregNoel> happens to all of us 19:42:53 that module does a lot of "clever" stuff like that 19:42:58 I knew there had to be some bad magic there. 19:43:05 kind of impressive, actually, but it makes things kinda fragile 19:43:07 <GregNoel> so something special is saved, but what, exactly? 19:43:11 Cause it does work for me on a daily basis. 19:43:49 <GregNoel> The sconsign command doesn't slow it (which may not be a surprise) 19:43:49 Here's its custom build info class: 19:43:51 class SConfBuildInfo(SCons.Node.FS.FileBuildInfo): 19:43:51 """ 19:43:51 Special build info for targets of configure tests. Additional members 19:43:51 are result (did the builder succeed last time?) and string, which 19:43:51 contains messages of the original build phase. 19:43:52 """ 19:43:56 result = None # -> 0/None -> no error, != 0 error 19:43:58 string = None # the stdout / stderr output when building the target 19:44:00 def set_build_result(self, result, string): 19:44:02 self.result = result 19:44:04 self.string = string 19:44:50 so there's a little magic at work 19:44:59 <GregNoel> Major magic 19:45:06 yeah, sconsign doesn't know anything about all this 19:45:09 it should 19:45:19 i sense a new issue being opened... 19:45:27 <GregNoel> you bet 19:45:58 But that makes me wonder what this bug is about. It should work as is. 19:46:00 working it 19:47:28 I bet Configure isn't overriding the main signature method hard enough. 19:48:09 it dates back to 0.96.95, might have been fixed since then 19:48:29 sounds like this needs research 19:48:48 <GregNoel> OK, you? 19:49:13 ok 19:49:23 I just tried it, it works for me on the trunk. 19:49:23 i can go with gary's classification: 1.x, p3, me 19:49:33 <GregNoel> done 19:49:40 or do we just call it WORKSFORME and let it get re-opened if necessary? 19:50:03 <GregNoel> hmmm..... Yes, I like that better 19:50:08 <GregNoel> close issues if we can 19:50:16 Well, it works for me, on Ubuntu, python2.5. But if you guys trust me that much... :-) 19:50:25 oh, but we do! 19:50:36 <GregNoel> Gary, will you close it with that comment? Tell him to reopen it if it's still a problem. 19:50:46 OK. 19:50:47 done 19:50:58 1580: 19:51:16 1.x p3 rob 19:51:16 <GregNoel> rob 19:51:25 <GregNoel> works for me 19:51:44 1597: 1.x p3 bill 19:52:25 <GregNoel> yes, works; thanks, Bill, for volunteering 19:52:21 1604: 1.x p4 greg 19:52:51 <GregNoel> 1604, yes, add it to my list 19:53:47 1545: 1.x p2 greg? 19:54:04 <GregNoel> p2? Hmmm, ok 19:54:28 I'd recommend p3 but p2 is ok 19:54:39 i was going from the spreadsheet, i'm okay with p3 19:55:00 <GregNoel> It's not a hard job, but finding all the places will take some shaking out. 19:55:05 yes 19:55:50 <GregNoel> ok, I'm good with p2 19:56:10 <GregNoel> I think that ends this spreadsheet and I need to go 19:56:19 Good work all! 19:56:34 very good, thanks 19:56:35 Can you guys do next wk at the same time? 19:56:37 <GregNoel> see you guys next week? What time? 19:56:52 good for me 19:56:56 <GregNoel> This time would be better for me (19h00) 19:57:03 okay, let's go with it 19:57:14 hmm, i was hoping David would make it too 19:57:24 oh, well, we made good progress 19:57:26 This worked well for me. See you then! Maybe David next week? 19:57:31 <GregNoel> yes 19:57:42 all right, next week, same bat time, same bat station 19:57:50 <GregNoel> OK, I've got the logs, so I'll update that tommorrow 19:57:55 great. Who's entering the data into tigris? 19:58:05 <GregNoel> You guys 19:58:14 OK, I did it last time & it was easy. 19:58:16 you did it last week, i'll take it this week? 19:58:24 OK, your turn then. 19:58:32 thanks! 19:58:41 * GregNoel attacked by rugrats, gotta go! 19:58:43 'night all