1. SCons
  2. Core
  3. SCons

Wiki

Clone wiki

SCons / BugParty / IrcLog2008-08-04

19:21:47  <sgk_> 2124: 
19:22:16  <sgk_> brandon, it's the open filehandles thing? 
19:22:21  <Azverkan>     yeah 
19:22:23  <[GregNoel](GregNoel)>     I think he's getting the errors because the shims aren't being installed 
19:22:33  <Azverkan>     I reproduced the bug localluy 
19:22:43  <[GregNoel](GregNoel)>     With the shims? 
19:22:57  <Azverkan>     I haven't figured out whether I can reproduce the problem with Microsoft's CL.EXE yet 
19:23:03  <Azverkan>     err wihtout 
19:23:10  <Azverkan>     it might be something specific to what they are doing 
19:23:17  <sgk_> if it's without then it's a known issue 
19:23:26  <sgk_> and we need to fix the other bug where the message isn't getting printed 
19:23:49  <[GregNoel](GregNoel)>     Yes, the fact that he found the bug suggests that he should have been getting the message. 
19:23:51  <sgk_> i also think we probably need to make those shims Yet Another configurable item 
19:24:15  <[GregNoel](GregNoel)>     Why would you want to turn them off? 
19:24:20  <sgk_> Carsten's message on the ML about compiler output getting lost when redirected is because of the shims, I think 
19:24:28  <Azverkan>     I think it might only applicable to applications that [CreateFile](CreateFile) with non default filesharing params, but I need to narrow it down more yet 
19:25:21  <sgk_> but that would need to be confirmed 
19:25:30  <Azverkan>     whats the "shim"? 
19:25:33  <sgk_> that aside, where does it leave us for 2124? 
19:25:45  <[GregNoel](GregNoel)>     Hmmm...  Sounds to me like Brandon should be working on it 
19:25:58  <Azverkan>     I think its a real bug that probably won't get fixed in the 1.0.x timeframe 
19:25:59  <sgk_> we replace the Python open() call (<ins>builtin</ins>.open()) with our own wrapper that calls the real underlying open() 
19:26:00  <[GregNoel](GregNoel)>     shim ==> code in Platform/win32 that replaces open() and file() 
19:26:19  <sgk_> and then calls the win32 api to supress the filehandle inheritance 
19:27:19  <Azverkan>     can verify while we move on to other bugs 
19:27:29  <sgk_> okay, cool 
19:27:29  <[GregNoel](GregNoel)>     2124: Brandon, research? 
19:27:34  <sgk_> done 
19:27:51  <[GregNoel](GregNoel)>     We'll see if he comes up with anything before we go 
19:27:50  <sgk_> 2155: 
19:28:10  <sgk_> okay 
19:28:18  <[GregNoel](GregNoel)>     2155: consensus invalid 
19:28:23  <sgk_> 2155:  sounds like we agree to deprecate it 
19:29:07  <sgk_> guess that would mean holding this open for now ut changing it to track the eventual deprecation? 
19:29:12  <[GregNoel](GregNoel)>     OK, I'll untangle it; probably put a note in 2126 
19:29:29  <sgk_> cool, thanks 
19:29:26  <Azverkan>     is there a naming convention for internal vs external functions? 
19:29:33  <sgk_> not really 
19:29:53  <sgk_> the closest we have is that, in general, public functions have [CamelCase](CamelCase) spelling 
19:30:02  <[GregNoel](GregNoel)>     External usually have caps; internal not, but it's not completely reliable 
19:30:07  <sgk_> but there are plenty of things like subst() that have gained currency 
19:30:34  *      sgk_ wishes he had [GregNoel](GregNoel)'s gift for brevity 
19:30:53  <sgk_> i always take a lot of extra words to say things... 
19:30:56  *      [GregNoel](GregNoel) can't type fast, so he learned to be brief 
19:31:28  <sgk_> okay, 2155:  [GregNoel](GregNoel) to handle 
19:31:46  <[GregNoel](GregNoel)>     done 
19:31:57  <[GregNoel](GregNoel)>     2157? 
19:32:20  <sgk_> 2157:  fixing ^C in configure context in general is WONTFIX 
19:32:42  <sgk_> but making sure we don't record an interrupt as a cached failure is... 1.x p3? 
19:33:07  <[GregNoel](GregNoel)>     it's probably the same thing; a partial store of something 
19:33:49  <[GregNoel](GregNoel)>     I'll go with looking at it again as 1.x p3 
19:33:58  <sgk_> done 
19:34:16  <[GregNoel](GregNoel)>     2158? 
19:34:19  <sgk_> 2158: consensus 1.0.1 p4 anybody 
19:34:25  <[GregNoel](GregNoel)>     done 
19:34:39  <[GregNoel](GregNoel)>     so who? 
19:34:41  <sgk_> OMG we got through the current issues 
19:34:54  <[GregNoel](GregNoel)>     only four! 
19:35:03  <sgk_> and in only half an hour! 
19:35:14  <[GregNoel](GregNoel)>     true. 
19:35:20  *      sgk_ is feeling full of piss and vinegar 
19:35:29  <sgk_> on to the backlog! 
19:35:36  <[GregNoel](GregNoel)>     Should we assign someone when we re-triage? 
19:35:44  <sgk_> for 2158? 
19:35:47  <[GregNoel](GregNoel)>     yes 
19:36:05  <sgk_> who? 
19:36:32  <[GregNoel](GregNoel)>     Anyone: you, me, Gary, Brandon.  It's one line of code! 
19:36:42  <sgk_> me 
19:36:45  <[GregNoel](GregNoel)>     done 
19:37:58  <sgk_> well, let's at least kill those last 2006H2 bugs 
19:38:24  <[GregNoel](GregNoel)>     yes, just getting set up.  My network is slow today 
19:38:32  <[GregNoel](GregNoel)>     1490 
19:39:04  <Azverkan>     (output from scons -j50 is funny btw, currently inserts other command lines between the \r and \n on windows so you get all sorts of weird line termination and interleaved output) 
19:39:40  <sgk_> Brandon:  like with -j3 you can get ccclll...eeexxxeee ///fffooo 
19:39:42  <sgk_> that sort of thing? 
19:40:08  <Azverkan>     all sorts of extended ascii characters and whatnot 
19:40:28  <Azverkan>     depends what code page you have setup 
19:40:29  <sgk_> yeah, we need something that slurps up that stuff and controls the output 
19:40:45  <sgk_> 1490:  boy, do i wish we had an installer guru 
19:40:53  <[GregNoel](GregNoel)>     that's the colorizer issue 
19:41:02  <[GregNoel](GregNoel)>     1490, yes 
19:41:09  <sgk_> agreed re: colorizer 
19:41:26  <[GregNoel](GregNoel)>     As I said in the spreadsheet, I have no clue. 
19:41:42  <[GregNoel](GregNoel)>     you two have to work it out 
19:42:03  <sgk_> let's say 1.x p3 (my favorite) 
19:42:15  <sgk_> since we have no one immediate to research it 
19:42:33  <[GregNoel](GregNoel)>     Brandon?  If you agree, I'll go along 
19:42:38  <sgk_> i can try to recruit someone who knows about installation 
19:43:03  <sgk_> i have a couple of people in mind who might be suitable 
19:43:16  <[GregNoel](GregNoel)>     And an AIX guru, and a Solaris guru, and ... 
19:43:33  <sgk_> agreed 
19:44:04  <Azverkan>     1490 we are using the stock installer 
19:44:21  <Azverkan>     I'd say wontfix with use a different install method? 
19:44:38  <sgk_> right, and it's getting clearer that stock distutils installer isn't up what we need 
19:45:10  <[GregNoel](GregNoel)>     I like wontfix; it closes an issue 
19:45:12  <sgk_> how about 1.x p3 me 
19:45:33  <[GregNoel](GregNoel)>     OK, although you've got too much work 
19:45:34  <sgk_> eh, i don't like losing input for when someone takes a look at it 
19:45:44  <sgk_> yeah, but i'd take this with the intent of passing it off 
19:45:48  <sgk_> once i find someone 
19:45:51  <[GregNoel](GregNoel)>     works for me 
19:45:59  <sgk_> done 
19:46:03  <[GregNoel](GregNoel)>     1.x p3, then 
19:46:39  <sgk_> 1500:  sounds like it might be a generic path interpretation issue 
19:46:44  <sgk_> research, who? 
19:46:50  <[GregNoel](GregNoel)>     I think 1500 is a mingw v. native issue 
19:47:34  <[GregNoel](GregNoel)>     (Man, I commented on this so long ago, I've almost completely forgotten what it was about.) 
19:47:46  <sgk_> hell, you're right 
19:48:04  <[GregNoel](GregNoel)>     always {;-} 
19:48:01  <sgk_> sounds like Cygwin, though, not MinGW 
19:48:27  <sgk_> of course!  we have to have at least *one* constant to rely on in the group...  :-) 
19:48:50  <Azverkan>     my guess is that he has both msys and cygwin in his path 
19:49:07  <sgk_> normal node-to-string translation is spitting them out as windows path 
19:49:11  <Azverkan>     before -mno-cygwin was added to GCC there was a lot of that 
19:49:12  <sgk_> yeah, i think Brandon is right 
19:49:16  <[GregNoel](GregNoel)>     Since I don't use them, I don't know the difference between MSYS and Cygwin 
19:49:32  <Azverkan>     cygwin tries to give windows a posix style interface 
19:49:36  <sgk_> Cygwin is evil evil evil 
19:49:38  <Azverkan>     msys tries to make native ports 
19:49:58  <[GregNoel](GregNoel)>     er, I wasn't asking, I was stating 
19:50:19  <Azverkan>     cygwin wouldn't be so bad if they were allowed to link against the now mostly unsupported posix runtime for the NT kernel 
19:50:57  <sgk_> or if it had just used the "liberal in what you accept" philosophy and not gagged on native Windows path names 
19:51:11  <sgk_> and if it didn't LIE about being able to support case sensitivity... 
19:51:36  <sgk_> sorry, Cygwin has caused SCons more than a few hassles over the years 
19:51:52  <sgk_> so my battle scars make me a little over-sensitive 
19:52:26  <Azverkan>     re the previous bug 
19:52:29  <sgk_> back to 1500:  WONTFIX 
19:52:34  <Azverkan>     the shim is the <ins>builtin</ins>.file = _scons_file? 
19:52:38  <sgk_> yes 
19:53:24  <[GregNoel](GregNoel)>     wontfix closes an issue, so I like the idea 
19:53:44  <sgk_> i can go ahead and write up the text and close it 
19:53:50  <[GregNoel](GregNoel)>     OK, I appreciate it 
19:53:59  <Azverkan>     either wontfix or invalid with a request for a test case 
19:54:24  <[GregNoel](GregNoel)>     1502? 
19:54:33  <sgk_> 1502:  i like 1.x p4 
19:54:51  <[GregNoel](GregNoel)>     done 
19:54:55  <sgk_> and eventually doing it with [GetOption](GetOption)() 
19:55:03  <[GregNoel](GregNoel)>     (that was almost too easy) 
19:55:20  <sgk_> statistically, *some* of them have to be... 
19:55:28  <sgk_> 1504:  research, [VisualStudio](VisualStudio), stevenknight 
19:55:35  <[GregNoel](GregNoel)>     done 
19:55:51  <sgk_> 1508:  research, [VisualStudio](VisualStudio), stevenknight 
19:55:51  <[GregNoel](GregNoel)>     although I think those are 'anytime' 
19:56:00  <sgk_> in practice, true 
19:56:16  <[GregNoel](GregNoel)>     in database, true 
19:56:24  <sgk_> if you want to change all [VisualStudio](VisualStudio) tags to anytime, that'd be fine with me 
19:56:29  <[GregNoel](GregNoel)>     1508, done 
19:56:33  <sgk_> or else I'll do it one of these days 
19:56:41  <[GregNoel](GregNoel)>     exactly 
19:56:50  <Azverkan>     re 2124: can reproduce with or without the shim installed 
19:56:55  <sgk_> ("in database, true" ?  not sure i get the reference, but i like the concept) 
19:57:16  <sgk_> Brandon:  yow, 2124 with the shim is definitely something uninvestigated 
19:57:25  <[GregNoel](GregNoel)>     ('anytime' sorts just after 1.0.x right now, so it'll show up on your radar) 
19:57:46  <sgk_> okay 
19:58:11  <sgk_> that makes 2124 relatively high priority in my book 
19:58:46  <[GregNoel](GregNoel)>     Brandon, can you look at it?  And report back next week? 
19:58:54  <Azverkan>     yep 
19:59:09  <[GregNoel](GregNoel)>     ok, let's just change the assignment, then 
19:59:18  <[GregNoel](GregNoel)>     er, owner 
19:59:40  <sgk_> cool; many thanks 
19:59:57  <[GregNoel](GregNoel)>     1516, speaking of colorization... 
20:00:01  <sgk_> 1516:  colorizer! 
20:00:02  <sgk_> yes 
20:01:09  <[GregNoel](GregNoel)>     If I understand it better now, process spawning === convert to subprocess. 
20:01:39  <[GregNoel](GregNoel)>     As for colorization, I'm color-blind, so it's never made any sense to me. 
20:01:40  <sgk_> yes 
20:01:57  <Azverkan>     for greg s/color/blink/ 
20:02:22  <[GregNoel](GregNoel)>     So, do we have an issue to convert to subprocess? 
20:02:34  <sgk_> no, there's a subprocess compatibility module now 
20:02:45  <[GregNoel](GregNoel)>     Azverkan, blink-blind?  I don't get it. 
20:03:08  <Azverkan>     blinkizer 
20:03:16  <sgk_> but hooking it into subprocess isn't really the right place 
20:03:16  <[GregNoel](GregNoel)>     Yes, but do we already have an issue to hang it on, or should we open one? 
20:03:44  <Azverkan>     I believe its also the same part of the multiprocessor buffering 
20:03:52  <Azverkan>     related to that is 
20:04:03  <[GregNoel](GregNoel)>     Ah, you're confused.  This issue is to colorize the output; is there another issue for the subprocess conversion? 
20:04:10  <Azverkan>     some sort of mechanism for attaching to spawn output 
20:04:29  <[GregNoel](GregNoel)>     Yes, apparently subprocess allows that. 
20:04:53  <[GregNoel](GregNoel)>     It doesn't pump the output to a process, but you can capture the output. 
20:05:56  <Azverkan>     subprocess just makes popen less buggy on windows?  afaik it doesn't offer anything above and beyond regular popen? 
20:06:32  <sgk_> i think you're right re: underlying capability 
20:06:42  <Azverkan>     once you have the text stream out of popen or subprocess you need some way to install listeners 
20:06:51  <sgk_> it's more about providing a more consistent, unified cross-platform interface 
20:07:37  <sgk_> 1516:  since I've already been in the middle of it, keep it with me 
20:07:48  <sgk_> 2.x ? 
20:07:54  <[GregNoel](GregNoel)>     ok, both sides? 
20:08:12  <Azverkan>     agree with 2.x 
20:08:15  <sgk_> sure 
20:08:21  <[GregNoel](GregNoel)>     colorize+subprocess? 
20:08:33  <[GregNoel](GregNoel)>     2.x, what priority? 
20:08:41  <sgk_> p3 
20:08:53  <[GregNoel](GregNoel)>     done 
20:09:10  <[GregNoel](GregNoel)>     Wow, that finishes this spreadsheet... 
20:09:12  <sgk_> okay, i have to get going 
20:09:18  <sgk_> same time next week? 
20:09:26  <[GregNoel](GregNoel)>     I should as well; cul? 
20:09:34  <Azverkan>     sure, later 
20:09:37  <[GregNoel](GregNoel)>     er, hang on... 
20:10:13  <[GregNoel](GregNoel)>     When shall we all meet again? 
20:10:13  <[GregNoel](GregNoel)>     In thunder, lightning, or in rain? 
20:10:13  <[GregNoel](GregNoel)>     Where the place? ...  same time next week? 
20:10:13  <[GregNoel](GregNoel)>     Should we change the default time to this time? 
20:10:30  <sgk_> you 20h00 PDT? 
20:10:44  <sgk_> you mean 20h00 pdt? 
20:10:49  <[GregNoel](GregNoel)>     19h00, same as you 
20:11:34  <sgk_> yes, we've been using this time regularly enough it should become the default 
20:11:48  <[GregNoel](GregNoel)>     OK, I'll take care of that, too. 
20:11:52  <Azverkan>     quick question regarding the shim 
20:11:57  <[GregNoel](GregNoel)>     sure 
20:12:00  <Azverkan>     it opens with the inherit flag on 
20:12:07  <Azverkan>     and some amount of time later it toggles it off? 
20:12:29  <[GregNoel](GregNoel)>     ah, a race... 
20:13:13  <sgk_> yes, since as far as i know, there isn't a way to open it with the inherit flag off 
20:13:38  <sgk_> if i'm wrong, that'd be great 
20:13:42  <[GregNoel](GregNoel)>     it could explain the symptoms, even with the shim enabled. 
20:13:40  <Azverkan>     have to research that more and my wife wants to eat now, so later 
20:13:56  <[GregNoel](GregNoel)>     OK, cul 
20:14:10  <sgk_> true 
20:14:20  <sgk_> later 
20:14:23  <sgk_> thanks, guys 
20:14:29  <[GregNoel](GregNoel)>     g'night 

Updated