Wiki

Clone wiki

SCons / BugParty / IrcLog2008-10-15

17:56:53 * garyo-home (n=chatzill@209-6-158-38.c3-0.smr-ubr3.sbo-smr.ma.cable.rcn.com) has joined #scons 17:59:18 <GregNoel> Hey, Gary... 17:59:25 <garyo-home> Hi, Greg. 18:00:00 <GregNoel> Steven's not here yet; anyone else here for the bug party? 17:59:48 <garyo-home> I gave a talk on SCons last weekend. Just need to upload it to the wiki. 18:00:27 <GregNoel> Yes, you mentioned it last time. The wiki sounds like a good place. 18:01:06 * stevenknight (n=stevenkn@72.14.224.1) has joined #scons 18:01:14 <GregNoel> Speaking of the devil... 18:01:28 <garyo-home> Hi, Steven. 18:01:51 <garyo-home> I just uploaded my SCons talk to the wiki. http://scons.org/wiki/GaryOberbrunner?action=AttachFile&do=get&target=SCons-talk-2008.pdf 18:01:53 <stevenknight> hey 18:03:01 <garyo-home> So, how about getting going? 18:03:11 <GregNoel> I'll look at it afterward; yes, let's go. 18:03:22 <GregNoel> 2220 18:03:44 <stevenknight> sorry, hang on, still getting set up 18:04:12 <GregNoel> Apparently it works in 0.98.something, but not since 18:04:06 <garyo-home> close as invalid, make new issue w/ test case & description, then retriage? 18:04:40 <GregNoel> Yes, but I'd feel better if we settled the timeframe now. 18:04:56 <stevenknight> agree w/garyo-home re: invalid and new issue 18:05:05 <garyo-home> IMHO it depends on how serious the actual issue is. 18:05:20 <garyo-home> If it only happens with nested builders, then 2.x p4 etc. 18:05:43 <GregNoel> No, my example used nothing but VariantDir 18:06:09 <garyo-home> Ah yes, I see that one now. 18:06:13 <stevenknight> okay, if greg's example is pure variantdir and a 0.98 regression 18:06:24 <stevenknight> then either 1.x 18:06:39 <stevenknight> or 1.2 (with likelihood of falling off the plate depending on priority relative to other stuff) 18:06:44 <stevenknight> my name on it 18:07:09 <garyo-home> ok, then 1.x p3? 1.2 is impossible at this point IMHO. 18:07:11 <GregNoel> Then I'd suggest 1.3 or 1.x 18:07:44 <stevenknight> 1.x p3 is fine with me 18:07:50 <GregNoel> ok, done 18:08:14 <garyo-home> 2225: 1.x Jim p3? 18:08:21 <GregNoel> 2226, yes 18:08:43 <stevenknight> i have 2225 next... 18:08:43 <GregNoel> oops, 2225 18:08:44 <garyo-home> 2225 or 2226, Greg? 18:09:32 <GregNoel> The new spreadsheet from Google allows me to set the font larger; you bet I'm going to use that next time so I can read the thing. 18:09:15 <garyo-home> consensus? 18:09:38 <GregNoel> yes, consensus 18:09:48 <stevenknight> 2225 yes consensus 18:09:55 <stevenknight> glad to hear from jim... 18:10:07 <GregNoel> Yes, we've missed him 18:10:21 <garyo-home> re: jim, yes! 18:10:00 <garyo-home> 2226: wontfix 18:10:49 <stevenknight> 2226: greg, agree w/WONTFIX? 18:10:49 <GregNoel> I can see the use case for 2226: do it once for the common case, rather than in dribs and drabs. 18:11:10 <GregNoel> I'd like to see a better patch, for sure 18:11:14 <jtc> For 2225, I agree with Jim's comment on the spreadsheet that in the long term we need to look at quoting more holistically. In particular, I think we need to look if we can defer quoting until just before spawning the command. Most make implementations will avoid spawning a subshell if there are no shell metacharacters. It is difficult for scons to do the same if everything has been quoted (although I suppose a de-quoter could be written). 18:11:14 <garyo-home> But it only speeds up the initial build. After that it's cached anyway. 18:11:33 <garyo-home> Hi jtc! 18:11:45 <stevenknight> jtc: hi! agree w/what you said re: quoting 18:12:09 <garyo-home> Yes, definitely. We just need to keep all cmd lines as lists or CLVars etc. until the last minute. 18:12:10 <jtc> At IDE, we experienced problems with high -jN, as the subshells caused us to run against the per-user process limit twice as fast as we would have liked. 18:12:24 <GregNoel> Yes, I hope the subprocess module will allow us to clarify it. 18:12:34 <garyo-home> Good point, Greg. 18:12:48 <stevenknight> subprocess will help 18:13:02 <stevenknight> but you still have command pipelines and redirection that will have to be detected 18:13:17 <garyo-home> Sure, but if it's all in one place it's not that hard. 18:13:18 <GregNoel> I don't see Jim here, but we've talked aabout how to do the quoting internally; maybe we should jointly prepare a proposal. 18:13:35 <stevenknight> that sounds good 18:13:40 <garyo-home> That would be great. Discuss on ML. 18:13:47 <GregNoel> yes 18:13:51 <stevenknight> on to 2226? 18:13:58 <stevenknight> or back to it 18:14:17 <garyo-home> 2226: we have too much to do already; this is a trivial addition even if it were fully formed. 18:14:36 <GregNoel> for 2225, I'm only proposing that we give it the 'toolchain' keyword so we look at it again when we're revamping the toolchain. 18:14:49 <GregNoel> sigh, 2226, 18:14:49 <stevenknight> 2225: toolchain++ 18:14:55 <garyo-home> ok I guess, but I don't think it has anything to do w/ that really. 18:15:08 <stevenknight> sorry, i'm confused 18:15:43 <GregNoel> My eyes can't resolve 5 v. 6 so I keep typing the wrong one. Sorry. 18:15:56 <garyo-home> no prob. 18:16:03 <stevenknight> 2226: not clear if David's trying to make it easier to configure or more efficient (one compilation vs. multiple) 18:16:17 <garyo-home> I thought it was just efficiency. 18:16:19 <GregNoel> a combination of both 18:16:41 <stevenknight> i think you give up too much by putting everything into one compilation 18:16:45 <GregNoel> trying a dozen things at once is much faster if they all work; 18:17:00 <GregNoel> if not, you fall back to testing one at a time 18:17:32 <stevenknight> within the call? or do you have to write that logic in your SConscript? 18:17:40 <garyo-home> imho, put it on the wiki as a custom SConf test. 18:18:06 <garyo-home> I think David's point is that on most platforms all the funcs will be there, so you just want a quick sanity check. 18:18:25 <GregNoel> yes 18:18:27 <jtc> As the maintainer of the autotools build for ACE/TAO (which may be the largest single autotools using project), I'm not sure if that holds. 18:18:46 <garyo-home> jtc: I agree, just pointing out his rationale. 18:19:27 <jtc> For example, it has feature tests for traditional UNIX and traditional MS Windows APIs. You typically won't find both. 18:19:55 <garyo-home> right, that's why I think the whole idea's a bit questionable. 18:19:57 <GregNoel> Uh, no, you'd combine the IX tests or the DOS tests not both in the same flow 18:20:49 <GregNoel> But I'm willing to go along; we're taking too long on this. 18:20:51 <garyo-home> greg: true, but you're still not going to test for every* DOS func you call, so it's not really here nor there. 18:21:23 <garyo-home> How about 2227? 18:21:34 <stevenknight> yeah, let's move on -- this really seems like an unnecessary optimization 18:21:40 <stevenknight> 2227: 18:22:07 <garyo-home> 2227 is the first time I've ever heard anyone say "ParseConfig works fine on windows" 18:22:11 <garyo-home> :-/ 18:22:17 <stevenknight> consensus 2.x p3 ? 18:22:29 <garyo-home> ok w/ me 18:22:31 <GregNoel> ok, 18:22:46 <GregNoel> maybe we'll change our mind by then 18:23:14 <GregNoel> 2228, consensus? 18:23:27 <garyo-home> yep 18:23:30 <stevenknight> 2228: done 18:23:42 <GregNoel> 2229, consensus? 18:23:43 <garyo-home> 2229, ditto 18:23:59 <stevenknight> 2229: consensus 18:23:58 <GregNoel> 2230 18:24:19 <garyo-home> I'd like it, but maybe makes more sense for 2.x than 1.x. 18:24:34 <GregNoel> 2230, I'll go with Steven 18:24:43 <stevenknight> 2230: okay, 2.x or anytime? 18:25:11 <garyo-home> I think it's worth 2.x rather than anytime 18:25:16 <stevenknight> okay, 2.x 18:25:33 <GregNoel> you suggested 1.x in the spreadsheet? 18:26:10 <stevenknight> after more thought i'm agreewing w/garyo's suggestion that 2.x is more realistic 18:26:21 <GregNoel> ok, I'll go with 2.x. what priority? 18:26:45 <garyo-home> p3 or p4, steven's preference 18:26:49 <stevenknight> p3 18:26:51 <GregNoel> done 18:26:54 <garyo-home> 2231: more warn opts 18:27:08 <GregNoel> Er, not quite. 18:27:41 <GregNoel> The idea is that a user may not know which new deprecation flags have been added 18:28:03 <GregNoel> so they just use --warn=all-deprecated and they get all of them 18:28:18 <stevenknight> that's what --warn=deprecated is supposed to do 18:28:35 <stevenknight> the hierarchy means that it will match all of the subclassed DeprecatedWarnings classes 18:28:51 <GregNoel> No, the first deprecation stage is a warning that is off by default 18:29:05 <stevenknight> ??? 18:29:18 <GregNoel> We didn't use it in this last round, but that's the way it's supposed to be. 18:29:12 <stevenknight> if you specify --warn=deprecated that means "on" 18:29:21 <stevenknight> and it will (or should) match your explicit settings 18:29:26 <stevenknight> before it looks at the defaults 18:29:49 <GregNoel> So there's a state you can't specify on the command line? 18:29:49 <stevenknight> didn't use what in this last round? 18:29:58 <stevenknight> what state? 18:30:34 <GregNoel> Three states, just like it says in the issue: warning off by default, warning on by default, warning not suppressible. 18:31:11 <GregNoel> And three master control options: turn on options normally off, use the default, and turn off suppressible options. 18:31:35 <stevenknight> sorry, i really don't get it -- i don't think we should ever have warnings that aren't suppressible 18:32:10 <GregNoel> OK, you will get screams of outrage when users are suddenly forced to upgrade. 18:32:28 <stevenknight> ??? 18:32:59 <GregNoel> Yes, there will be a set of people who always run with --warn=no-deprecated 18:33:25 <GregNoel> They will be rudely surprised when they are suddenly forced to change their scripts 18:32:04 <stevenknight> maybe we should take this off line so you can explain it to me 18:33:11 <garyo-home> I think offlining this is a good idea. 18:33:50 <GregNoel> I'll agree to that, so retriage the issue next time? 18:33:57 <garyo-home> ok 18:34:03 <stevenknight> ok 18:34:14 <garyo-home> (w/ additional info in the ticket) 18:34:22 <GregNoel> ok 18:34:35 <GregNoel> 2232, I checked, it's fixed, I'll close it 18:34:46 <garyo-home> great 18:34:50 <stevenknight> cool 18:35:09 <garyo-home> 2233: I'll reply to OP and get details 18:35:20 <GregNoel> good, I'll leave it to you 18:35:32 <stevenknight> 2233: good, thanks 18:35:39 <GregNoel> retriage next time then? 18:35:59 <garyo-home> sure, depending on reply. 18:36:18 <GregNoel> done 18:37:21 <GregNoel> 2234, consensus for anytime? I don't like making an actual defect an open-ended issue. 18:38:26 <garyo-home> It seems really easy; 1.x should be OK. 18:38:35 <stevenknight> 2234: 1.x is fine with me 18:38:50 <GregNoel> what priority? 18:39:05 <stevenknight> p3 18:39:11 <GregNoel> done 18:39:27 <GregNoel> 2235 18:39:48 <garyo-home> definitely make code agree w/ doc here 18:40:17 <stevenknight> 2235: agree 18:40:36 <GregNoel> OK, I'll run regressions and see what it catches. As far as I know, there's only one test that does anything with them. 18:40:48 <garyo-home> (hmm, my kids are still awake, it's 9:40 on a school night... grr) 18:41:02 <stevenknight> garyo-home: i feel your pain... 18:41:07 <garyo-home> greg: regressions = good idea. 18:41:22 <GregNoel> OK, anytime is acceptable? 18:41:24 <stevenknight> okay, done with current issues 18:41:29 <jtc> the curse of parenthood... 18:41:33 <stevenknight> anytime is fine with me -- or research 18:41:53 <garyo-home> anytime 18:41:54 <GregNoel> anytime 18:42:00 <GregNoel> done 18:42:01 <stevenknight> done 18:42:22 <GregNoel> One question before we go on... 18:42:50 <garyo-home> Hey, allofasudden I can edit 2005h2 and never could before (using the regular link). Maybe it's the new google docs upgrade. 18:42:55 <garyo-home> yes, greg? 18:43:17 <GregNoel> Steven mentioned that he's normally getting off the shuttle at 18h30 or thereabouts; should we move the time earlier by a half-hour? 18:43:45 <garyo-home> That makes it a little harder for me. 18:45:04 <GregNoel> I suspected that, but it took us 45 min to clear tonight's issues; we need more than a half-hour if we're meeting at 18h00 18:45:25 <stevenknight> i could see about shifting my schedule on the nights we have these 18:45:33 <stevenknight> so happened that i worked from home today 18:45:54 <GregNoel> Always a good schedule... {;-} 18:46:10 <garyo-home> I could probably do it at 18h30 though, since it's only every other week. 18:47:02 <GregNoel> Is that better for you, Steven? 18:47:15 <stevenknight> probably a little 18:47:29 <stevenknight> if i take the shuttle on those nights, it gets in right about 18h30 18:47:42 <stevenknight> but i could find a wifi cafe and join pretty shortly after 18:47:43 <GregNoel> OK, I'll post it that way; Steven, will you keep us informed if it has to move? 18:47:50 <stevenknight> will do 18:48:09 <GregNoel> OK, onward. 18:48:18 <garyo-home> Wait, I meant to say half hour earlier would be ok -- but half hour later is better for me, is that what we just agreed on? 18:48:33 <stevenknight> right, half hour later, 18h30 PDT, 21h30 EDT 18:48:40 <garyo-home> ok, thanks! 18:48:51 <stevenknight> cool 18:49:03 <stevenknight> shall we make some headway on 2005h2? 18:49:39 <garyo-home> 1230: consensus worksforme 18:49:40 <GregNoel> worksforme 18:49:45 <stevenknight> done 18:49:47 <stevenknight> 1235: 18:49:54 <garyo-home> consensus fixed 18:50:03 <stevenknight> i might have already closed it 18:50:06 <stevenknight> 1241: 18:50:14 <garyo-home> invalid, I'm ok w/ that 18:50:20 <stevenknight> 1241: invalid 18:50:21 <GregNoel> done 18:50:21 <stevenknight> done 18:50:40 <garyo-home> 1244: let me research that. Looks like some good stuff might be in there. 18:50:47 <stevenknight> oh, damn -- that's right, i couldn't edit this for a while, either 18:50:58 <stevenknight> 1244: research, garyo, done 18:51:10 <GregNoel> done 18:52:08 <GregNoel> 1249? 18:52:23 <stevenknight> 1249: i like your suggestion: ludwig, research 18:52:34 <garyo-home> Could Mkdir just succeed if target exists, and also create intermediate dirs? 18:52:54 <GregNoel> It does. 18:53:12 <GregNoel> but it then tries to make the intermediate directory 18:53:19 <GregNoel> and fails 18:53:26 <garyo-home> ... but why doesn't that Mkdir succeed also? 18:53:52 <GregNoel> os.mkdir fails: directory already exists 18:54:00 <garyo-home> It should trap that error and ignore it. 18:54:27 <GregNoel> Yes, it checks, but it checks before the other Mkdir creates the directory 18:55:00 <stevenknight> needs some more research, then? or greg, do you feel you've characterized it sufficiently to identify the right fix? 18:55:21 <GregNoel> Sure, but then, I think Ludwig should do it 18:55:33 <garyo-home> I can fix it in 10 minutes including test. 18:55:37 <garyo-home> Just give it to me. 18:55:40 <GregNoel> done 18:55:55 <stevenknight> done 18:56:15 <garyo-home> (But I'm only going to fix the proximate cause, not whatever Ludwig's patch is about.) 18:56:23 <GregNoel> Just make sure it still fails if it's a file (or whatnot) that's preventing the creation 18:56:26 <stevenknight> that works for me 18:56:33 <garyo-home> Good point, Greg. 18:56:41 <GregNoel> or file permissions, or anything else. 18:56:52 <garyo-home> Right, no problem. 18:56:59 <GregNoel> Ludwig's patch clears the cache when the directory is created 18:57:24 <garyo-home> Ah, right, so the next Mkdir gets the test right. 18:57:37 <garyo-home> ok let's move on 18:57:38 <GregNoel> you mean wrong 18:57:43 <garyo-home> :-) 18:58:01 <stevenknight> 1253: 18:58:20 <stevenknight> greg, did you reproduce with current scons? or with 0.96.91? 18:58:36 <GregNoel> current, with the .sconsign he provided 18:58:59 <stevenknight> ah 18:59:10 <stevenknight> i'm inclined to either WORKSFORME or RESEARCH, then 18:59:16 <stevenknight> the .sconsign file would have changed since then 18:59:21 <stevenknight> so it's not surprising that we can't handle it 18:59:30 <GregNoel> but we should detect that, yes? 18:59:51 <stevenknight> we do. that's why we print the warning 19:00:07 <GregNoel> Er, I think it's a fatal error now 19:00:08 <stevenknight> if we didn't detect it, you'd get a stack trace 19:00:43 <GregNoel> It's been a while, and I'm not positive, but I think it did give a stack trace 19:01:23 <stevenknight> okay, sounds like research me 19:01:26 <GregNoel> done 19:01:38 <stevenknight> note re: making sure it doesn't stack trace 19:01:54 <GregNoel> done 19:02:28 <GregNoel> 1260 19:02:37 <garyo-home> 1260: probably moot due to recent fortran work 19:02:56 <GregNoel> Probably, but I think David should check it out 19:03:04 <garyo-home> David should check, agreed. 19:03:04 <stevenknight> what greg said 19:03:10 <stevenknight> research, David? 19:03:16 <GregNoel> research? 19:03:27 <garyo-home> ok 19:03:30 <GregNoel> done 19:03:54 <GregNoel> 1261, whatever you guys decide 19:04:23 <garyo-home> Interesting. I hadn't seen that. Do we have cygwin platform support now? 19:04:35 <stevenknight> kinda sorta 19:04:49 <stevenknight> never had a real cygwin expert do a thorough job with it 19:05:11 <stevenknight> we do have places where we account for cygwin differences 19:05:35 <stevenknight> (especially its really annoying characteristic of lying about case sensitivity) 19:05:40 <garyo-home> I don't think there's anything like this patch in tools now, and it looks pretty OK. I'm inclined to take it seriously. 19:05:46 <stevenknight> agree 19:06:05 <GregNoel> (Three years old, remember) 19:06:17 <garyo-home> It's basically a gcc-lookalike with some tweaks. 19:06:51 <stevenknight> sounds reasonable 19:06:55 <stevenknight> i can take it 19:06:56 <garyo-home> Greg: if we have this in, it'll help us remember what to do on cygwin in the toolchain stuff. 19:07:02 <garyo-home> Steven: great 19:07:07 <stevenknight> what time frame? 19:07:11 <garyo-home> 2.x? 19:07:16 <stevenknight> that sounds right 19:07:17 <garyo-home> p3? 19:07:21 <stevenknight> yes 19:07:24 <GregNoel> done 19:07:27 <stevenknight> add a cygwin keyword? 19:07:43 <GregNoel> or 'toolchain'? 19:07:56 <stevenknight> or both? 19:07:59 <garyo-home> either or both, ok w/ me 19:08:14 <GregNoel> Steven, your choice 19:08:25 <stevenknight> i was thinking both might be handy in case someone tries to tackle cygwin before toolchain (or vice versa) 19:08:36 <GregNoel> done 19:09:15 <GregNoel> 1263? 19:10:38 <stevenknight> needs to be reproduced, it's been a while 19:10:42 <stevenknight> i bet it's been fixed since then 19:11:06 <stevenknight> better if someone else has time, but i'll take it (research) if no one else can 19:11:43 <GregNoel> Trivial to reproduce; it's using glob.glob() instead of Glob(), so it's in the "wrong" directory the second time through. 19:12:35 <stevenknight> ah! 19:12:35 <garyo-home> That's probably right... 19:12:44 <garyo-home> (actually os.listdir, but same thing) 19:12:51 <stevenknight> close it out, then, with reference to Glob() ? 19:13:14 <GregNoel> yup, that's what I said in the spreadsheet... 19:13:16 <garyo-home> I think so. OP can reopen if desired (ok, it's 3 yrs old, they won't...) 19:13:35 <GregNoel> done 19:13:46 <GregNoel> 1268 19:14:08 <stevenknight> ah, okay, i stopped scrolling down on the spreadsheet 19:14:09 <stevenknight> 1268: 19:14:36 <stevenknight> agree w/greg: research, Jim 19:15:07 <garyo-home> ok, but my quick look says this patch couldn't hurt. 19:15:23 <jtc> gotta go folks; I'll try to make the next bug party ... 19:15:30 <GregNoel> Let Jim decide 19:15:32 <garyo-home> thanks, J.T.! 19:15:35 <stevenknight> thanks, jtc 19:15:38 <GregNoel> We'll look for you 19:15:43 <GregNoel> the more the merrier! 19:16:12 <garyo-home> re: let jim decide, ok. 19:16:36 <GregNoel> done 19:16:56 * jtc has quit ("Quit") 19:17:30 <GregNoel> 1276 19:17:49 <stevenknight> 1276: kind of hairy and architectural 19:17:57 <stevenknight> i'm probably the logical assignee, unless someone else wants it 19:17:59 <garyo-home> 1276: I guess Greg's ssheet comment is right. 19:18:00 <stevenknight> agree w/future 19:18:03 <garyo-home> future. 19:18:12 <GregNoel> what priority? 19:18:25 <stevenknight> p2 sounds right 19:18:32 <GregNoel> done 19:18:33 <stevenknight> shorter sk: agree w/greg :-) 19:19:31 <GregNoel> Huh? Where did I say that? 19:19:47 <stevenknight> no, i was poking fun at myself 19:20:03 <stevenknight> the summary of my previous long-windedness was: I agree w/greg 19:20:19 <GregNoel> ah 19:20:33 <stevenknight> 1281: 19:20:56 <stevenknight> agree we need a Java guru 19:21:20 <GregNoel> I had no clue with this one, and re-reading it, I still don't 19:21:24 <stevenknight> if we had one, what priority / timeframe 19:21:35 <stevenknight> arbitrary: 2.x p3 ? 19:21:42 <garyo-home> whatever 19:21:58 <stevenknight> that lets us defer until 1) someone pops up; 2) we get to it eventually 19:21:58 <GregNoel> OK, I'll go with that 19:22:31 <garyo-home> 1282: is dup of 1268 19:22:41 <garyo-home> sorry I mean 12385 is dup 19:22:51 <GregNoel> keep trying 19:22:56 <garyo-home> sorry, 3rd try: 1285 is dup of 1268 19:23:01 <garyo-home> yes, that one was right. 19:23:04 <garyo-home> :-) 19:23:31 <stevenknight> okay, dup 1268 19:23:40 <GregNoel> done 19:23:43 <garyo-home> I just marked it as dup. 19:24:50 <garyo-home> 1287: 19:25:10 <stevenknight> yow, patch that's been hanging around way too long 19:25:17 <garyo-home> I think copying the attributes is the right idea. 19:25:25 <stevenknight> yeah, sounds exactly right 19:25:32 <stevenknight> shouldn't be too hard to cook up a test case 19:25:43 <stevenknight> give it to me, p2, 1.2 or 1.x 19:25:44 <stevenknight> ? 19:25:53 <GregNoel> your choice 19:25:57 <garyo-home> your choice 19:26:17 <stevenknight> 1.2 19:26:18 <GregNoel> done 19:26:48 <GregNoel> 1290 19:26:48 <garyo-home> 1290: I think scons used to write the .sconsign incrementally 19:26:58 <garyo-home> maybe it does again now? 19:27:22 <garyo-home> Anyway we are better about signal handling so it rarely fails to update the .sconsign 19:27:36 <stevenknight> yeah, i think we could WONTFIX it 19:27:37 <garyo-home> I think it's invalid due to better signal handling now 19:27:45 <garyo-home> WONTFIX is ok 19:27:48 <GregNoel> done 19:27:52 <stevenknight> done 19:28:00 <stevenknight> 1293: 19:28:05 <GregNoel> 1293 19:28:41 <stevenknight> probably research, me again... :-/ 19:28:57 <garyo-home> or me, at least I could try to repro it quickly 19:29:07 <garyo-home> I have a D drive 19:29:23 <stevenknight> garyo, go for it 19:29:41 <garyo-home> ok 19:29:44 <GregNoel> done (Steven has too many research issues anyway) 19:29:53 <stevenknight> agreed 19:29:59 <GregNoel> 1211 19:30:11 <GregNoel> (and this is the last one in this spreadsheet) 19:30:28 <stevenknight> (yay!) 19:30:52 <stevenknight> old, seems to be fixed, don't spend time on it, just WORKSFORME and invite re-opening if that's hasty 19:31:05 <garyo-home> agree w/ both of you. 19:31:06 <GregNoel> worksforme! 19:31:34 <stevenknight> excellent work tonight, gents 19:31:39 <garyo-home> yes 19:31:41 <GregNoel> OK, we've settled on 17h00 in two weeks? 19:31:50 <stevenknight> 18h30 ? 19:31:54 <GregNoel> oops, 17h30? 19:32:08 <garyo-home> I think it was 18h30 PDT 19:32:10 <stevenknight> 18h30 ? 19:32:56 <GregNoel> Uh, I'll have to scroll back, ah, ok, I was arguing for 17h30, but I guess I kept mistyping it 19:32:58 * jrandall (n=jim@bas1-london14-1096624847.dsl.bell.ca) has joined #scons 19:33:03 <GregNoel> Hi, Jim 19:33:14 <garyo-home> ok, see you then guys. Hi, Jim! 19:33:30 <GregNoel> We assigned you a bunch of issues, Jim 19:33:32 <jrandall> hello - I seem to be somewhat late to the party 19:33:37 <stevenknight> okay, see you later, gary 19:33:41 <GregNoel> Yes, just ending 19:33:43 <garyo-home> l8r 19:33:48 <stevenknight> hey jim -- better late than never, though 19:33:49 <GregNoel> g'night 19:34:31 <jrandall> I'll check the log for the summary. More quoting stuff? 19:34:38 <stevenknight> yep 19:35:17 * garyo-home has quit ("ChatZilla 0.9.83 [Firefox 3.0.3/2008092417]") 19:35:38 <GregNoel> And there's a comment in the spreadsheet about a possible strategy to deal with quoting 19:35:52 <jrandall> Nice - I'll check that out right now 19:36:23 <jrandall> Current issues sheet? 19:37:14 <GregNoel> no, 2005h2 issue 1268 19:37:20 <jrandall> Is the intent of scons to expose the host quoting scheme? 19:37:39 <GregNoel> I'd argue not 19:38:02 <GregNoel> in fact, I'd suggest using shlex to crack incoming strings 19:39:12 <jrandall> The "quoting model" was kind of the fundamental question I ran into. 19:39:20 <jrandall> And was unable to decide which I'd prefer. 19:39:47 <jrandall> The "host quoting scheme" seemed to be what I'd naturally assume, but that's tough on the project independance 19:40:44 <GregNoel> Yes, but there are so many incompatible schemes on DOS, so I'd prefer to pick one that's consistent and just go with it 19:41:13 <GregNoel> not to mention that Python has built-in support for Bourne-style shell quoting 19:42:25 <jrandall> So suggestion would be to use bourne-style shell quoting for all scons commands? 19:42:47 <jrandall> or one scheme, whatever it may be, on all host platforms? 19:43:32 <GregNoel> we do something similar for ParseConfig; the input is assumed to be GNU-style flags, which are placed in the right variables so they're usually "translated" to the native format 19:44:27 <GregNoel> not sure I understand your distinction between SCons commands and one scheme for all 19:45:06 <jrandall> wasn't trying to distinguish - rather tired, and not speaking well :) 19:46:23 <GregNoel> Yeah, I understand that---I've been getting up at 2am (PDT) the past few days, so this is well past my bedtime... 19:46:24 <jrandall> two approaches seem to be "crack into tokens, we control the quoting", or "foist quoting onto the host platform, never try to bust up strings" 19:46:36 <jrandall> That's a bit on the early side :) 19:47:11 <jrandall> The latter seems less fraught with peril, and probably more compatible with existing practice, but not as nice cross-platform 19:47:50 <GregNoel> It's a long story, but the short is that it's 110+ during the days right now, so we agreed that our contractors could get here at a ghastly hour to start work. 19:48:13 <jrandall> ouch. 19:48:36 <jrandall> I'm not sure exactly what 110+ translates to in celsius, but I'm pretty sure it's damn hot :) 19:49:17 <GregNoel> "less fraught with peril" is my motivation. I think consistent and predictable is the win here. 19:49:52 <GregNoel> over 44 degrees 19:50:29 <jrandall> Aye. There seems to be an endless supply of quoting issues. As per the comment on 1268, that pretty much summarizes what needs to be able to be done if subst_list is oging to work 19:50:56 <jrandall> and if we can't crack into a list of tokens like that, then almost have to not rely on subst_list 19:50:59 <GregNoel> Yeah, I saw your comment about that, but I haven't looked at it yet 19:51:46 <jrandall> Part of while tempfilemunge is such a bug magnet is that it's built on subst_list, which likes to bust strings on spaces. 19:52:46 <jrandall> so it either has to be able to understand quoting or not be used in tempfilemunge. Some other quoting problems in a similar vein 19:53:21 <jrandall> it == subst_list in previous sentence :) 19:53:50 <GregNoel> ambiguity, thy name is pronoun... 19:55:57 <GregNoel> Anyway, it looks like I have to go; can you drop me a line about this? I'd like to see if we can come up with a spec to describe it, particularly as we make the move to subprocess, which will make all of the quoting issues go critical again. 19:56:19 <jrandall> Sure thing. see you 19:57:19 <GregNoel> (Subprocess takes a list of strings, which are assumed to be pre-quoted, and figures out how to get them run. If we can figure out how to create that list of strings, we win big.) 19:57:34 <GregNoel> Yes, the wife is calling... cul. 19:57:46 <jrandall> Hrm, a good reason to stick with the list approach. 19:57:47 <jrandall> see you. 19:58:39 * jrandall (n=jim@bas1-london14-1096624847.dsl.bell.ca) has left #scons

Updated