IrcLog2008 07 14

18:53:15 * GregNoel is no longer marked as being away 18:58:55 * stevenknight (n=[email protected]) has joined #scons 19:00:04 <GregNoel> Hi, Steven. Gary has said he would likely be late; anybody else here for the bug party? 19:00:21 i don't see Bill, and he's the other stalwart 19:00:57 <GregNoel> And only you and I commented in the spreadsheet, and you didn't finish. 19:00:59 i'm just getting into the Current Issues spreadsheet -- I'm taking th late shuttle home tonight 19:01:08 right, just catching up 19:01:17 the existing comments were mine from last week 19:02:03 * garyo-home (n=[email protected]) has joined #scons 19:02:13 <GregNoel> Gary's not that late, after all 19:02:24 Hi Greg. 19:02:44 Hi, Steven. 19:02:55 <GregNoel> Hey, Gary. You said you would be late. 19:03:23 Snuck out just in time, or mostly. 19:03:43 <GregNoel> I think Steven is in a different window, updating the current issues spreadsheet; he should be back soon. 19:03:36 hey gary 19:03:39 Hi 19:03:43 how'd your release go last week? 19:03:56 GregNoel's ESP ++ 19:04:12 Release went great. I haven't got a lot of time for scons these days due to things at work. 19:04:31 We're growing the company, got new investors, new CEO... lots of new & exciting stuff 19:04:39 but it takes up all my time & then some. 19:04:43 <GregNoel> The disadvantage of working for a living... 19:04:55 ...says the retired Unix guru. 19:05:03 <GregNoel> {;-} 19:05:03 :-) 19:05:27 So anyway, that's all in apology for the fact that I haven't touched the spreadsheets. 19:05:20 well, shall we make as good use of the time as we can, then? 19:05:39 Yes, let's dive in. Current issues first? 19:05:42 i might disconnect briefly in ~10 minutes, i have to transfer shuttles 19:05:45 yes current issues 19:05:47 <GregNoel> 2124 19:06:12 1.x p3 me 19:06:20 ok w/ me. 19:06:23 <GregNoel> I admit a traceback is unfriendly, and something should be done about that, but the problem is that ... 19:06:34 parts of the VS revamp will try to clean up some general windows issues 19:06:40 <GregNoel> he's really using a different name for the file. 19:07:08 <GregNoel> With that said, 1.x p3 makes as much sense as anything. 19:07:23 okay, let's go with it 19:07:27 <GregNoel> done 19:07:29 * bdbaddog (n=[email protected]) has joined #scons 19:07:36 <GregNoel> Hey, Bill. 19:07:39 2121 has come up a few times on the list, right? 19:07:41 Hi Bill. 19:07:51 Bill! 19:08:29 <GregNoel> Yes, and I think there may be dups, but I couldn't find them. 19:08:41 what is there about the confusing VariantDir feature that hasn't come up a few times on the list? 19:08:46 The patch seems reasonable on the face of it. 19:09:05 <GregNoel> (patch?) 19:09:19 212 has a patch and a test. 19:09:23 sorry 2121. 19:10:07 <GregNoel> Ah, it looks like that came along after I commented. 19:10:09 Anyway, I agree w/ you guys on 1.x p2. 19:10:38 <GregNoel> Yes, 1.x p2 is even stronger with a patch to work from. 19:10:50 yeah, 1.x p2 -- the patch looks good (haven't looked at the test case) and should be rewarded 19:10:59 ok, good. 19:11:01 <GregNoel> done 19:11:54 2122 is a way not to have to use src_builder iiuc? 19:12:04 right, essentially 19:12:13 let you add new src_builders dynamically 19:12:13 <GregNoel> I don't know if this is the best API, but I agree that it something should be done. 19:12:45 and with some supported API so everyone doesn't have to cut-and-paste all the obj_builder stuff that's initialized in Tool/init.py 19:12:43 <GregNoel> I published the long-promised requirements for better messages earlier today; that has a comment about this issue. 19:13:10 sounds good; i'll take a look when we're done 19:13:12 func name is maybe not perfect but yes something like this is good. 19:13:50 any objections to sticking with 1.x p3? 19:13:50 <GregNoel> I think better messages and this are indirectly related, so fixing one will have an impact on both 19:14:04 But since it's an enhancement, I'd say low pri for 1.x (p3 max) or else later. 19:14:27 <GregNoel> My suggestion is the same as better messages, and I don't remember what that was assigned. 19:14:53 1458? 19:15:08 um, nope. 19:15:09 greg, what was the thread from earlier today re: better messages? 19:15:12 you have me intrigued now 19:15:28 <GregNoel> wiki BetterMessages 19:15:35 okay 19:15:36 2123: 19:15:51 consensus 1.x p2 ? 19:15:58 <GregNoel> fine with me 19:16:09 who? 19:16:11 ok. I can probably do it. 19:16:42 It looks pretty easy. 19:16:42 okay, thanks -- just added your name to the spreadsheet 19:16:45 2125: 19:17:30 <GregNoel> 2122: http://scons.org/wiki/BetterErrorMessages 19:17:41 2125: if Tools inherited from a base class, they wouldn't have to implement exists(). 19:18:09 have to switch buses, might drop momentarily 19:18:46 ... and if they were subclasses it'd be easy to see what's a Tool. 19:18:51 <GregNoel> Tools are not classes; they're modules (i.e., imported) 19:19:34 Yeah (though there are other types, but classes aren't among them). I guess we can't really change that. 19:20:01 <GregNoel> Not easily; there's also the backward-compatible issue. 19:20:22 A module can inherit stuff, but doing that just to avoid writing 'return True' seems overkill. 19:20:52 I think this bug is making a mountain out of a molehill; should be 2.x low pri if anything. 19:21:25 Greg, what you say in the ssheet is spot on. 19:21:42 <GregNoel> I agree; it's overkill. That's why I suggested wontfix. 19:21:52 I agree, wontfix. 19:22:13 <GregNoel> If Steven makes it back without dropping, we can have a consensus. 19:22:38 * sgk_ (n=[email protected]) has joined #scons 19:22:46 .. and here he is now. 19:22:50 <sgk_> I'm back -- thought I was still connected but I guess not 19:22:54 <GregNoel> We'll probably be changing this interface with the toolchain stuff, but I'd like to leave it until then. 19:23:09 <sgk_> still on the exists() thing? 19:23:14 <GregNoel> Yes 19:23:11 Greg & I say "wontfix" 2125. 19:23:16 yes, exists(). 19:23:35 <sgk_> do new-style classes allow it to be treated like gary was suggested (re: subclassing)? 19:23:45 <sgk_> old-style classes definitely didn't 19:24:04 don't know 19:24:06 <GregNoel> I don't think so... 19:24:15 <sgk_> okay, well not terribly important 19:24:41 <GregNoel> do we have a consensus? 19:24:46 <sgk_> this was from a colleague lobbying me re: all the cut-and-paste "def exists(): return True" at the bottom of all the written modules 19:24:58 <sgk_> wontfix is fine with me 19:25:10 you can blame it on us. 19:25:22 <GregNoel> yeah, we're hardcore 19:25:34 <sgk_> lol 19:25:43 <GregNoel> 2126? 19:25:44 <sgk_> 2126 then: 19:26:11 <sgk_> no real strong feelings so far -- any reason not to leave it 1.x p4? 19:26:11 Having these as functions would be nice, I say 1.x p4 19:26:25 <sgk_> done 19:26:28 <sgk_> 2127: 19:26:40 <GregNoel> Moving to Python 2.2 would allow these to be written as simple names, 19:26:51 <GregNoel> but that would require waiting until 2.x 19:27:04 <sgk_> ah, that should be at least noted in the issue 19:27:12 <GregNoel> OK, wilco 19:27:15 <sgk_> i'll add a comment in the background here 19:28:18 <sgk_> 2127: 19:28:53 <GregNoel> 2127, I'd like to spin this off onto someone who has the background with all the variations. 19:29:23 I do, but even with that it's not clear what the right answer is. 19:29:23 <GregNoel> But who? I surely don't. 19:29:32 Good evening all. 19:29:52 If a user says RPATH=XXX, should we try to provide those semantics by jiggling other linker args? 19:29:57 <GregNoel> Hey, Bill... 19:30:01 Hi, Bill. 19:30:29 Greetings finally back from HI, and then OC. phew. 19:30:41 <GregNoel> Somehow, autoconf figures it out, since they support rpath, but ... 19:30:44 <sgk_> sounds like there's enough uncertainty that 2127 should either be a research for someone 19:30:55 <GregNoel> ... the complexity looks intimidating. 19:31:20 I'll be happy to research it. But at some point scons has to say "this compiler doesn't support RPATH (or not well enough)" and punt. 19:31:25 <GregNoel> Your research or my research? They're different. 19:31:21 <sgk_> or a 1.x-p3-and-reprioritize if "research" is too much of a backburner 19:31:40 <GregNoel> Ah, your research. 19:31:45 <sgk_> yours (i.e., should be investigated) 19:31:49 I have a bunch of Macs with different OSes, so I can at least poke them all. 19:31:49 <sgk_> heh 19:31:58 <sgk_> okay, garyo research 19:32:15 <GregNoel> My research takes priority over 1.0, i.e., research it now. 19:32:24 <sgk_> i think research should be Greg's interpretation (AIIU, investigate for reprioritization) 19:32:31 <sgk_> but in practice that doesn't seem how we're handling it 19:32:38 <sgk_> right 19:32:51 <GregNoel> but if Gary wants to do it, I'll let him have it. 19:33:04 <GregNoel> garyo research 19:33:05 (Hmm, do I have any research items? Not sure...) what I want is 1.x research (i.e. research as a priority) 19:33:05 <sgk_> okay, gary, research 19:33:30 <sgk_> that's kind of what I've morphed 1.x p3 into, mentally 19:33:34 but I'll get something done on it. 19:33:47 <GregNoel> no, research and 1.x are both milestones; can't change the names of the priorities. 19:33:51 <sgk_> I figure we're going to have a big reprioritization of 1.x issues at some point 19:33:57 <sgk_> to break them down into manageable chunks 19:34:04 <sgk_> cause there's just too much there right now 19:34:12 <GregNoel> You do have a talent for understatement {;-} 19:34:44 oh well, that just means there may be lots of 1.x's 19:34:59 (or we slip things til 2.0 of course) 19:35:16 <GregNoel> Aye, there's the slip, er, rub 19:35:21 anyway, 2128 is next... 19:35:29 <sgk_> maybe. we need to discuss releasing 1.0 (I think 0.98.5 has baked enough) 19:35:32 <GregNoel> 2128, David 19:35:45 <sgk_> and when/how to branch so there's a place for relevant dev work 19:35:54 <sgk_> 2128: david 19:36:04 2128 Includes doc patch, I say 1.0 or 1.0.x. 19:36:15 <GregNoel> True, but not quite yet; one issue later may need to be slipped in. 19:36:23 Steven: yes, it's getting to that point. 19:36:48 We can branch it any time and just merge things that need to go in. 19:37:00 <GregNoel> You're looking at 2129; no patch for 2128 19:37:10 <sgk_> 2128: 1.0 for the doc patch 19:37:19 <sgk_> ? i see an attachment to 2128 19:37:23 me too. 19:37:39 a trivial two-liner. 19:37:53 <sgk_> 2129 is another david Fortran thing, though 19:38:22 2129: wow, a patch which is just a test. 19:38:27 <sgk_> 2129: anyone, anytime (it's an added test) 19:38:29 <GregNoel> Yeah, but is it the doc or the implementation? 19:38:45 2128: doc. 2129: test for implementation. 19:39:02 <sgk_> no, greg's suggesting that although 2128 might "fix" the doc, 19:39:12 <GregNoel> OK, 2128 1.0 David, 2129 anytime 19:39:13 <sgk_> the doc might be right (the CPP variables should be in the command line) 19:39:16 <sgk_> and the code needs fixing 19:39:22 aha, I see. 19:39:57 <GregNoel> David either way. 19:40:00 We would need David to answer that. 19:40:07 * stevenknight has quit (Read error: 110 (Connection timed out)) 19:40:14 there goes Steven. 19:40:18 <sgk_> right, done 2128: david, 1.0, with a note about the doc-vs.-code 19:40:20 <GregNoel> Ah, we just lost Steven... 19:40:25 <sgk_> hey , where'd i go? 19:40:39 <GregNoel> vanished in to the AEther... 19:40:46 hm, my irc client said your connection timed out. 19:40:41 <sgk_> (that was the connection through the other bus timing out) 19:41:04 I see. 19:41:17 <sgk_> okay, 2129: anyone, anytime 19:41:28 <sgk_> 2130: 19:41:50 2130, doc license issues: can we satisfy them somehow, maybe a CC license of some kind? 19:42:03 <GregNoel> Have you figured out what he really wants? 19:42:06 That would let you still print the UG? 19:42:17 <sgk_> CC license would be the right thing, i suppose 19:42:30 <sgk_> this is probably a research, me to figure out how where to draw the line 19:42:38 <GregNoel> OK, works for me 19:42:44 <sgk_> yeah, they want to make the UG available on (e.g.) Debian 19:43:01 <sgk_> but it's copyright me, not the SCons Foundation, and it's unclear if they can legallly do it 19:43:03 <GregNoel> although getting it into 1.0 would be good 19:43:09 <sgk_> i'll sort it out 19:43:14 <GregNoel> OK, works for me 19:43:20 ok 19:43:28 <sgk_> just changed it to research (Greg's research) 19:43:48 <GregNoel> Ah, really? 19:44:04 <sgk_> well, i won't promise, but I do conceptually agree with it 19:44:26 ok, 2131 (glob needs to sort)? 19:44:28 <GregNoel> I thought I understood the initial request, but not since. 19:44:31 <sgk_> in practice, right now i'm prioritizing UG updates over research to get 1.0 out 19:44:59 <sgk_> 2131: is there any downside to making Glob() return a deterministic order? 19:45:02 <sgk_> i can't think of one 19:45:18 <GregNoel> glob.glob doesn't sort; why should Glob? 19:45:21 We should definitely sort it. 19:45:31 <sgk_> principle of least surprise 19:45:34 Who would want it in random order? 19:45:45 and you could use --random if you did... 19:45:53 <GregNoel> "least astonishment" yes, you're probably right. 19:46:00 <sgk_> having SCons rebuild things whenever it feels like it because you use Glob() seems really unehlpful 19:46:23 right, I think this should be 1.0.x p2. Easy and helpful. 19:46:33 gotta run. hey can someone look at my comments bug 243. I did some research and seems like a real bug where we thought it was doc bug before. 19:46:53 <GregNoel> later 19:46:56 <sgk_> okay, we'll try to look at 243 19:46:57 <sgk_> later 19:47:03 bye 19:47:14 <sgk_> 2131: 1.0.x p2? 19:47:36 <GregNoel> ok, I guess 19:47:55 fine w/ me. 19:48:12 <GregNoel> 2132 19:48:34 <sgk_> 2132: Ralf's fixes tend to be pretty good 19:48:40 <sgk_> i haven't lookat the code on this one yet, though 19:48:44 <sgk_> looked at 19:48:45 <GregNoel> sgk_, I'm pretty sure it was an earlier issue 19:48:55 <GregNoel> It uses subprocess 19:49:00 Can we use subprocess.Popen()? 19:49:16 <sgk_> should be able to, the compatibility layer has a subprocess module that works under 1.5.2 19:49:27 <GregNoel> we hope 19:50:01 <GregNoel> If we can't find the dup, I move for 1.0.x 19:50:16 <sgk_> agreed 19:50:22 <sgk_> 1.0.x... p3? 19:50:28 That early? OK I guess since there's a good patch. 19:50:28 <sgk_> or p2? 19:50:44 <GregNoel> yes, and if we find the dup, make it the same. 19:50:48 <sgk_> ~5 minutes until i leave the bus 19:51:04 <GregNoel> and we're not even out of the current issues... 19:51:09 <sgk_> i'll volunteer to hunt for the dup 19:51:12 <sgk_> so put my name on it 19:51:16 <GregNoel> ok, done 19:51:24 <sgk_> two weeks' worth 19:51:37 <GregNoel> but only five new ones 19:51:42 <sgk_> true 19:51:47 2133: invalid, or should we try to handle AddPostAction differently (no implicit dep on cmd)? 19:51:53 <sgk_> 2133: i think this case should work 19:52:01 <sgk_> it used to, and it doesn't seem unreasonable 19:52:09 <sgk_> ("should be made to work (again)" that is) 19:52:33 AddPostAction cmds don't really need to be dependencies anyway, so I agree. 19:52:39 <GregNoel> sounds like a hack... 19:52:49 <GregNoel> Hmmm... I think they do 19:52:53 No, because AddPostAction is not a builder. 19:53:01 <sgk_> agree w/gary 19:53:11 <GregNoel> think of a local command that JFCLs through the binary 19:53:12 <sgk_> plus it's easier to add an explicit Depends() if you really want that dependency 19:53:18 Builder cmds should get auto deps, but not Pre/Post actions. 19:53:20 <sgk_> than to shut it off 19:53:26 <GregNoel> the command should be rebuilt if it changes 19:53:52 <sgk_> hmm, Greg i do see your point -- SCM purity would require it 19:53:53 Greg: hm, I have to think about that. 19:54:11 <GregNoel> we're not going to settle this now; not enough time; resume here next time? 19:54:14 <sgk_> since you can't know the AddPostAction() is irrelevant 19:54:22 <sgk_> works for me 19:54:36 <GregNoel> OK, then, when next? 19:54:38 ok. Same time, same place, next week? 19:54:45 <sgk_> same time, etc. 19:54:53 <GregNoel> 19h00? or 17h00? 19:55:02 1900 is good for me, how about you? 19:55:07 <sgk_> 19h00 is fine with me 19:55:13 <GregNoel> fine with me 19:55:16 Greg: I'll do the data entry this week from your irc log 19:55:17 <sgk_> done 19:55:25 <sgk_> gary: thanks 19:55:43 <GregNoel> ok, although I have the time this week 19:55:50 <sgk_> i'll probably start a release@ thread re: really releasing 1.0 19:56:02 sgk_: I was just going to suggest that. 19:56:14 <GregNoel> good idea 19:56:28 Greg: thanks but I think I can handle it, gotta contribute somehow... 19:56:49 plus I'll be on vacation 23rd - 6th 19:56:50 <GregNoel> Personally, I'd rather you were editing the spreadsheets... 19:57:06 <sgk_> disconnect in < 15 seconds, later 19:57:08 OK, I agree. I'll make some time for that too. 19:57:19 <GregNoel> ok, later 19:57:24 * sgk_ has quit ("Leaving") 19:57:25 bye guys. 19:57:29 <GregNoel> cul 19:57:35 * garyo-home has quit ("ChatZilla 0.9.83 [Firefox 3.0/2008052906]") 19:57:37 * GregNoel has been marked as being away