IrcLog2008 05 19

16:42:23 * bdbaddog (n=[email protected]) has joined #scons 17:25:45 * stevenknight (n=[email protected]) has joined #scons 17:26:25 <GregoryNoel> That's three; where's Gary? 17:26:39 i think he said he might be late 17:26:42 putting the kids to bed 17:26:56 bill, you're not spinning tonight? 17:27:22 oh, wait, he's greyed out 17:27:30 who's the third? 17:28:17 * garyo-home (n=[email protected]) has joined #scons 17:28:25 <GregoryNoel> You, me, Bill, and there's Gary. 17:28:34 hi Gary 17:28:50 hi guys, I'm here for a little, then I'll have to put the kids to bed, then I'll be back. 17:28:57 <GregoryNoel> Anybody else here for the bug party? 17:30:09 <GregoryNoel> OK, the official start is here; shall we proceed? 17:30:16 let's go 17:30:24 starting with current... 17:30:25 2048 17:30:36 <GregoryNoel> In the overlaps? 17:31:00 the "Current issues" spreadsheet? 17:31:33 <GregoryNoel> We should do the ones left over from last time first; it overlaps with the current issues 17:31:49 okay, point me to the list/spreadsheet you want to work from 17:32:06 I guess that would be editlist2008, w/ 1874 first 17:32:14 <GregoryNoel> yep 17:32:35 go ahead, i've scrolled down 17:32:45 ok, 1874: hasn't this been fixed multiple times before? 17:33:10 * bdbaddog has quit ("Leaving.") 17:33:12 <GregoryNoel> Yes 17:33:14 I think I fixed it myself a long time ago. 17:33:27 Should've made a better test case I guess. 17:33:46 <GregoryNoel> Bill just left, but he was going to bring it up for discussion on the mailing list 17:34:03 <GregoryNoel> did it ever happen? I don't remember it. 17:34:13 Don't think so. 17:34:33 <GregoryNoel> And Steven is right that one size does not fit all. 17:34:59 I seem to remember that I fixed it by ignoring "suffixes" that were all numeric, that's probably why ".4g" fails that test. 17:35:21 Does this just have to be configurable? 17:35:31 <GregoryNoel> I don't see how 17:35:30 I think give it back to Bill and/or recategorize it as a doc issue 17:35:43 <GregoryNoel> I like doc issue 17:35:52 there could be a "force suffix" option or something? 17:36:06 <GregoryNoel> Hmmm 17:36:09 that's File("name-with.odd-suffix") 17:36:31 <GregoryNoel> good point 17:36:32 oh, you mean on the Builder 17:36:46 yeah I guess 17:37:00 <GregoryNoel> OK, a doc issue assigned to?? 17:37:14 I don't think it's just doc, is it? 17:37:35 <GregoryNoel> "If you don't like the suffix, use File()" 17:37:56 OK, I could live with that for now, but I'd like a better solution for 2.x 17:38:15 <GregoryNoel> I would, too 17:38:10 back to Bill to really discuss on the mailing list 17:38:34 and/or doc the File() workaround 17:38:30 <GregoryNoel> OK, I'll make it, what, research? 17:38:37 yeah, research 17:38:40 <GregoryNoel> done 17:38:45 <GregoryNoel> next? 17:38:54 1883 17:39:09 damn, i answered this in the other spreadsheet as well 17:39:14 my comment in editlist2008 is off 17:39:22 the last time we put it in we did have instaler issues 17:39:34 <GregoryNoel> What's up with Nathan? 17:39:37 but it was because distutils changed the location to the script/ subdirectory at the same time 17:39:45 I don't think it had anything to do with this App Paths thing 17:39:58 ??? 17:40:01 oh, GSoC? 17:40:13 <GregoryNoel> Yes, we talked about assigning it to him 17:40:39 ah, right 17:40:42 1883: what does the patch really do? I can't see it. 17:41:20 it adds an entry to the Windows registry 17:41:48 IIRC it ends up making it so you can execute scons.bat w/out having to have the directory in %PATH% 17:42:17 Nathan has been sending me status reports, but I've not been giving him adequate attention yet 17:42:20 That would be good; maybe it sets cmd.exe's AppPath or something I guess 17:42:22 so we could definitely assign it to him 17:42:45 I never use scons.bat, but I could do so for testing this. 17:42:47 <GregoryNoel> OK, what's his account? 17:43:15 i'll look it up 17:43:20 let's move on while i search 17:43:32 1925, then? 17:43:46 <GregoryNoel> Add it to the spreadsheet when you find it; I'll take care of it later 17:44:01 okay 17:44:07 1925: research, me 17:44:24 <GregoryNoel> done; next? 17:44:32 OK, but not for 1.0 though 17:44:53 definitely not 1.0 17:45:01 <GregoryNoel> probably 2.x 17:45:07 1958: Bill's volunteering, consensus research 17:45:29 <GregoryNoel> done 17:46:11 2000: I say 1.x but not the approach in the patch; should really figure it out. 17:46:28 <GregoryNoel> when? 17:46:56 Low priority, so could be 2.x as far as I care 17:47:02 <GregoryNoel> 1.x? or move to 2.x? 17:47:35 I say 2.x because it'll just slow down real work 17:47:37 agree w/Gary, 1.x, low priority... P4? 17:47:46 <GregoryNoel> p5 17:47:50 fair enough 17:47:57 <GregoryNoel> done; next? 17:47:59 do i hear p6? 17:48:08 <GregoryNoel> (no such!) 17:48:10 :-) 17:48:19 going once, twice.... sold! 17:48:43 ok, 2001? (remove max_drift) 17:48:58 2001: research 17:49:00 <GregoryNoel> Narrow use; 2.x 17:49:08 I don't use it, but it was meant to support NFS. 17:49:26 i'm inclined to give it back to Ken and let him lead a ML discussion to find out who's actually using it 17:49:29 if anyone 17:49:35 <GregoryNoel> And NFS now uses deltas, so it doesn't happen any more. 17:49:38 Actually Greg if it can give a 25% speedup, I say get rid of it sooner 17:49:50 speedup++ 17:50:01 1.x, p2? 17:50:12 * stevenknight agrees 17:50:22 <GregoryNoel> ok 17:50:44 2003 is weird 17:50:46 2003: agree w/Greg, wontfix 17:50:52 agree w/Gary, weird 17:51:07 wontfix 17:51:11 <GregoryNoel> done 17:51:37 <GregoryNoel> last one? 17:51:59 This will get fixed someday by Greg+Gary tool rewrite, but what about the near term? 17:52:39 We can't use his patch as is, people don't expect CCFLAGS to get clobbered. 17:53:06 <GregoryNoel> Maybe it should be set by c-common setup, whatever it's called. 17:53:34 <GregoryNoel> er, no, bad idea 17:53:53 hmm, i took a quick look last night and i think his narrow fix of having mingw reset $CCFLAGS solves his specific symptom nicely with little impact 17:54:02 <GregoryNoel> This happens because one compiler is configured and then another configured on top of it 17:54:15 agreed that's the larger issue 17:54:20 Right, but what if user sets CCFLAGS and then applies Tool('mingw')? 17:54:35 <GregoryNoel> He gets what he pays for. 17:54:44 they're no worse off than lots of other things that get set 17:54:52 $CCCOM, $CFLAGS, etc. 17:54:56 Hmm, OK I see your point. 17:55:10 all that has to wait until your tool rewrite 17:55:17 OK, 1.x then. 17:55:21 but we can make this one situation better in the meantime 17:55:26 <GregoryNoel> OK, what priority? 17:55:36 p2 or p3 17:55:58 p3, it's only that one tool in that one case 17:56:02 <GregoryNoel> done; on to the next spreadsheet 17:56:02 okay 17:56:24 Current Issues, right? 17:56:26 I like Ken's patch in 2048 17:56:44 gary, you think 1.0? 17:56:51 <GregoryNoel> destab 17:57:00 as in 0.98.5? 17:57:01 <GregoryNoel> destableizing 17:57:06 right, i'm worried abou stability on it 17:57:12 Look at the code; it only does changes that one case. But 1.x is fine w/ me. 17:57:16 <GregoryNoel> (ok, I still can't spell) 17:58:02 1953, my current bete noire... 17:58:03 <GregoryNoel> 1.x, what priority? 17:58:05 1.x, give it to me for integration 17:58:20 2048: p3? 17:58:24 p2, i agree that the patch is nice (modulo stability) 17:58:32 ok, p2 17:58:36 <GregoryNoel> 2048, ok 17:59:13 Can we put 1953 in 1.0? 17:59:35 At least to see if that fixes the problem? 17:59:37 that code looks safe enough to me 17:59:52 and there needs to be a 0.98.5 for other reasons anyway 17:59:55 1.0, p2 17:59:55 <GregoryNoel> it was off the top of my head; don't take it literally 18:00:07 right, but it's clear a problem 18:00:24 and you point to the right sort of solution, even if the code ends up a little different 18:00:26 and it has to be a threading thing because otherwise that error could not occur 18:00:28 <GregoryNoel> it still leaves the race, it just covers up the symptoms 18:00:36 Greg: that is true. 18:01:05 an ounce of image is worth a pound of performance... ;-) 18:01:33 I'll be back in a bit -- at least you're at the part of the spreadsheet where I did my homework now :-) 18:01:34 <GregoryNoel> Gary, have you tried it? 18:02:21 <GregoryNoel> Why don't we assign it to you for research; if it seems to kill the problem, we'll try it for 1.0 18:02:30 <GregoryNoel> And he's gone.... 18:02:37 yeah 18:02:40 research, me 18:03:00 <GregoryNoel> OK, I'm sure Gary will be willing to test it 18:03:11 agree about the underlying race for NodeInfo still being there; I'll add comments to that effect 18:03:24 in get_ninfo(), not just here 18:03:39 and/or in NodeInfo.init() or some such 18:03:48 <GregoryNoel> works for me 18:03:59 okay, looks like we covered the next set of overlaps 18:04:03 1874, 1883 18:04:05 <GregoryNoel> skipping the overlaps to 1967? 18:04:27 right 1967 18:04:30 consensus future 18:04:37 <GregoryNoel> done 18:04:40 do we need an assigee? 18:05:03 <GregoryNoel> for that far in the future? no, I don't think so. what priority? 18:05:18 leave it p3 18:05:24 <GregoryNoel> done;next? 18:05:41 skip 2000, 2001 18:05:49 <GregoryNoel> I'll take 2007 18:05:50 2007: 1.x, you 18:05:53 <GregoryNoel> done 18:06:19 2010: 2.x consensus 18:06:21 leave unassigned? 18:06:28 <GregoryNoel> yes to both 18:06:57 <GregoryNoel> I want to get a keyword for all of these so we can triage them further as a group. 18:07:06 <GregoryNoel> and assign them then 18:07:05 good idea 18:07:18 2014: i'm torn 18:07:39 <GregoryNoel> I don't understand why it's needed 18:08:04 right now we assume that no one else has corrupted the tree in between runs 18:08:08 not unreasonably 18:08:10 <GregoryNoel> either you trust the sig or you don't. 18:08:51 <GregoryNoel> if you don't trust it, always recalc, fine. 18:09:05 <GregoryNoel> but if you're going to try for optimizations, you have to trust it 18:09:21 hmm, i do see your point 18:09:26 <GregoryNoel> that's why Decider() has such a range of options 18:09:31 this was a bigger problem back when we were using build signatures 18:09:51 <GregoryNoel> but they're going away 18:10:00 and we could use signatures from the .sconsign file assuming no file corruption 18:10:01 <GregoryNoel> don't throw good effort after bad 18:10:40 <GregoryNoel> Uh, which file corrupted? .sconign? 18:11:06 no, you build 18:11:28 then someone corrupts your .obj file (or copies a trojan into it) 18:11:48 and we could see the .c file hasn't change, so we don't rebuild the .obj 18:12:06 but then use that corrupt .obj to link a .exe 18:12:13 so this verification would be 18:12:15 <GregoryNoel> The sig wouldn't match, oh, I see, 18:12:29 right 18:12:38 <GregoryNoel> The new sig wouldn't match, but the old one could. 18:12:40 <GregoryNoel> hmmm 18:13:22 right, it starts to use the .sconsign signatures as a weak bill-of-materials of sorts 18:14:01 before you use the built targets from last time, please make sure that you think they really do match what you thought you built 18:13:49 <GregoryNoel> How about a Decider(always-recalc)? 18:14:17 something like that 18:14:34 <GregoryNoel> I could understand that but I'd do it as a Decider() 18:14:42 give it to me, 1.x, p3 18:14:49 <GregoryNoel> done 18:15:13 if it fits in Decider I'll do it that way 18:15:58 hmm, looks like i'll be able to go beyond 6:30 tonight 18:16:12 <GregoryNoel> Keep pushing... 18:16:16 we're stuck in traffic 18:16:32 likely because of an accident... :-( 18:17:05 <GregoryNoel> (I'll tell you my stuck-in-traffic story some day) 18:16:36 <GregoryNoel> 2015 18:17:06 1.x, me, p3 18:17:47 <GregoryNoel> ok, getting that scan for the dir source really needs to be fixed 18:18:46 <GregoryNoel> 2016, consensus 18:19:22 yeah, 2.x 18:19:50 <GregoryNoel> 2020: this isn't tool config, why our plan? 18:20:31 <GregoryNoel> oops, screen update, nevermind 18:21:03 sorry, what are we on? 18:21:08 2016 is consensus 2.x, yes? 18:21:12 and I have 2018 next 18:21:52 <GregoryNoel> Yeah, the spreadsheet is giving me partial screen updates 18:22:06 okay 18:22:33 i think 2018 is pretty straightforward 18:22:43 1.x seems reasonable 18:22:45 <GregoryNoel> I'm pretty sure that blanks are compressed out of all cmd-STR variables 18:23:12 <GregoryNoel> But I've broken the case where I was doing it, so I'm not positive. 18:23:10 if you want to confirm that I'll support INVALID 18:23:24 <GregoryNoel> me, research? 18:23:45 done 18:24:15 <GregoryNoel> 2020, you, as specified, done 18:24:15 2020: me, 1.x, p...2? 18:24:49 2021: 1.x, anyone else's choice of priority 18:25:07 <GregoryNoel> These File/Dir conflicts are new; something started them. p2 is probbly OK 18:25:40 2021 is actually the --debug=time + --interactive bug, not File/Dir 18:26:18 <GregoryNoel> Yeah, I don't type fast enough 18:26:49 no problem, just want to make sure we're getting right info on the right bug 18:27:01 <GregoryNoel> 2021 p2 unless it's not simple, then p3 or p4 18:27:20 agreed 18:27:23 <GregoryNoel> done 18:27:55 2022: agree w/your plan, let David prioritize it relative to his time and other Fortran work 18:28:04 <GregoryNoel> 2023, you research for dup? 18:28:37 yes 18:28:49 <GregoryNoel> 2022, funny screen updates again, done 18:29:30 no problem 18:29:33 2029: 18:29:46 <GregoryNoel> not a lot of yacc users, 2.x? 18:29:59 i could go for that 18:30:13 <GregoryNoel> assign to Gary? 18:30:14 we can always move it up if there's a groundswell 18:30:20 <GregoryNoel> agreed 18:30:21 yes 18:30:24 <GregoryNoel> done 18:31:00 * GregoryNoel stays silent for 2036 18:31:05 2036: consensus 2.x p2 18:31:19 <GregoryNoel> done 18:31:23 i completely agree we're way overdue for a better way to do this 18:31:37 assign it to either me or you and we can work out a reasonable interface between us 18:31:45 i'd be happy to implement, though 18:32:08 <GregoryNoel> I don't like DESTDIR; too inflexible, but we can discuss that elsewhere 18:32:21 i think you're right 18:32:29 2037: TASK 18:32:36 it's not product code at all 18:33:03 let Sohail check it in himself and you (or anyone else) can hack on it as necessary 18:33:12 <GregoryNoel> done; I fiddled with it today 18:33:14 hi guys I'm back 18:33:24 hey there 18:33:33 we're up to 2041 on the "Current issues" spreadsheet 18:33:38 <GregoryNoel> and there's some nice test code in the attachment; Hi, Gary... 18:33:53 just in time for you to weigh in on it, it's an intelc.py thing 18:34:03 OK. yep, that sounds like mine. 18:34:32 okay, we got past the traffic slowdown, i think i've got another 5-10 minutes 18:34:52 Don't know if it has to be as complicated as that patch though; I might just add an option so user could specify if needed. 18:35:03 2041: assign to gary, 1.x, p...3? 18:35:05 Anyway assign it to me, 1.x, p2 or p3 18:35:11 <GregoryNoel> p3 18:35:24 <GregoryNoel> done 18:35:46 skip next (OVERLAP) 18:35:47 2043 18:35:50 2043 seems like a side project to me 18:35:56 2.x, p4 18:36:01 we have plenty of real work to do 18:36:05 OK, 2.x p4 18:36:21 <GregoryNoel> done; when shall we three meet again? 18:36:25 if they want to actually do the work i'd be okay with it going in earlier, too 18:36:36 <GregoryNoel> in lightning, thunder, or in rain? 18:36:38 2044: 18:36:45 1.x, p2 (if not p1)? 18:36:48 Yes, 2044 should be 1.x or earlier 18:36:54 i'd say 1.0 but it's potentially destabilizing 18:36:54 <GregoryNoel> uh, that was a left parenthesis... 18:36:57 UNC paths are important 18:37:20 if you want we could make it 1.0 and i could take a look at how bad it would be 18:37:36 i can always decide to push it back 18:37:42 <GregoryNoel> research? I don't want to commit to 1.0 18:37:58 I can run it here too. Let's just look at the code carefully before putting it in 1.0. 18:38:06 i'd prefere 1.x over research to make sure it stays on more visible lists 18:38:17 <GregoryNoel> 1.x p1? 18:38:19 research suggests "back burner" to me w.r.t. actually allocating time 18:38:19 Yes, don't make it research 18:38:24 yeah, 1.x p1 18:38:33 OK w/ that 18:38:39 <GregoryNoel> No, research means "figure this out and assign it to a milestone" 18:39:05 <GregoryNoel> Can we meet again tomorrow briefly if we don't finish today? 18:39:05 Greg: technically you're right but we're close to 1.0 now so there's not much research time left 18:39:09 i agree conceptually, but in practice I deal with 1.0 before research 18:39:28 Yes, I can do tomorrow night for a little while. I'll finish the spreadsheet too. 18:39:48 <GregoryNoel> Just for this spreadsheet 18:40:00 2046: consensus 1.x p4 18:40:03 i can do tomorrow 18:40:05 right, that's the last bit: 2046 to the end 18:40:27 okay, last few minutes for me 18:40:36 tomorrow night: 17:00 or 17:30? 18:40:52 17:00 is better for me I think 18:40:58 that's fine for me 18:41:00 <GregoryNoel> either is fine by me 18:41:04 17:00 18:41:09 ok, done, see you then 18:41:13 sounds good 18:41:15 many thanks 18:41:24 <GregoryNoel> 17h00 to use the standard, such as it is... 18:41:34 17h00... :-) 18:41:38 right. 18:41:39 <GregoryNoel> 2047? 18:41:57 That's the one that a user was complaining about, right? 18:42:09 How about warning instead of erroring? 18:42:11 yeah 18:42:29 <GregoryNoel> I'll buy a warning 18:42:36 1.0? 18:42:42 <GregoryNoel> hmmm 18:43:02 As long as adding the warning and keeping going is easy, then 1.0, else 1.x. 18:43:05 i'm more comfortable with 1.x, but this is pretty annoying 18:43:26 me, 1.0, p2 18:43:32 <GregoryNoel> 1.x p1; if he finishes early, we can reconsider 18:43:33 Put it in for 1.0 but if it gets tricky then reschedule for 1.x 18:43:36 if it looks risky i'll push it out 18:43:44 agreed 18:43:53 <GregoryNoel> which? 18:43:59 coming up to the bus stop, catch you guys tomorrow 18:44:01 1.0 18:44:13 ok, have a good night Steven! 18:44:17 2047: 1.0, p2 18:44:18 <GregoryNoel> Let's pick up here; cul 18:44:18 l8r 18:44:22 * stevenknight has quit ("Leaving") 18:44:24 (pun not intended) 18:44:47 ok Greg, I'll see you tomorrow as well. 18:44:54 <GregoryNoel> OK, cul 18:45:02 bye