**** BEGIN LOGGING AT Sun Apr 01 03:00:01 2018 Apr 01 04:10:09 occasionally Apr 01 15:22:32 suppose I have a native recipe that installs 2 GB, and is needed by many (~20) recipes. is there any way to install it to one place, rather than 20 different recipe sysroots, to save space? Apr 01 15:22:44 perhaps add it to the toolchain itself? Apr 01 17:57:35 * armpit working on morty build failure... Apr 01 21:05:16 laplante: assuming there are no hardcoded paths in it, all the copies will be hardlinks Apr 01 21:05:41 armpit: good luck! Apr 01 21:06:03 RP: thanks! Khem got back to me earlier today and pointed that out. I had misunderstood what 'du' was doing wrt hard links Apr 01 21:07:34 RP, its really odd. glibc-local is being installed then removed during the sdk creating Apr 01 21:07:47 armpit: :/ Apr 01 21:38:07 So does Yocto provide a modern embedded OS for my old Zaurus SL-5500 ? :) Apr 02 00:13:44 heh, i think i have an SL-5500 around the house somewhere. Apr 02 00:13:50 sharp took down all their SDKs at one point. Apr 02 00:14:36 seebs: I think I know why your syscall() test case succeeded but that build failed Apr 02 00:14:42 ooh Apr 02 00:15:36 The build was calling syscall() during the shared-library initialization (before main()), which must have been before pseudo shared library initialization Apr 02 00:15:59 hmm. maybe, but i *thought* pseudo initialized on first call to a wrapper. Apr 02 00:16:09 ... come to think of it, no, i think there's a couple cases where we try to do it earlier. Apr 02 00:16:25 No I check, pseudo does it on library initalization Apr 02 00:16:36 ... i had apparently forgotten this. Apr 02 00:17:28 ahh, yeah, back in _libpseudo_init(). which is, it turns out, not *sufficiently* reliably early, which is why all the wrappers have to check for initialization anyway. Apr 02 00:17:40 we probably don't strictly need to do it, even, but it seems like a good plan. Apr 02 00:18:16 I think you want to.... it could be a pain if it tried to initialize after threading is started Apr 02 00:18:22 Yeah. Apr 02 00:18:28 I mean, someone will probably eventually hit that anyway. Apr 02 00:18:41 I sorta wanna go study the cross-localedef failure and see if I can find out where the issue was. Apr 02 00:18:52 Because my guess is it's an actual bug in cross-localedef, just one which is usually-harmless. Apr 02 00:19:08 Although it's *possible* that they're using one of the rare calls which actually promises not to set errno unless there's an error. Apr 02 00:19:13 Ya, I wasn't paying any attention to that one :) Apr 02 00:20:03 The immediate cause was that the reworked path computation had an error in the initial logic for determining the starting state of the "what kind of thing are we looking at" stat buf. Apr 02 00:20:14 Which could cause it to call lstat("", &buf), which would set errno. Apr 02 00:20:29 Otherwise, you only get errno set by that code if you're resolving a path which actually contains a nonexistent file. Apr 02 00:22:07 one of the places i sent an application to had a thing saying "oh, and if you have an existing open source project, you can use that instead of doing this little programming exercise"... Apr 02 00:22:20 but it turns out they want something under 2k lines of code so they have any hope of jumping in and understanding it usefully. and that seems fair. Apr 02 00:22:56 ... I did, however, have one person tell me that they were just skipping the technical side of interviewing entirely because they knew what pseudo was. Apr 02 00:23:08 Ya... pseudo is not something I would necessarily want for an application example Apr 02 00:23:17 It is not pretty code. Apr 02 00:23:23 Really? I don't think that's fair, because what's their goal in understanding? Don't they just want to see how you work? Apr 02 00:23:26 I try to fix it up, but there's a lot of mandatory cruft. Apr 02 00:23:37 No... it necessarily has to do very not-pretty things Apr 02 00:23:38 Well, they wanted to be able to, say, ask questions about the thing, and see whether my answers made sense. Apr 02 00:23:42 It all depends on the job, really. I'd love to see somebody apply that wrote something similar to pseudo. Apr 02 00:24:01 And with something like pseudo, none of their people would necessarily have the background to know whether my answers were sensible. Apr 02 00:24:01 Oh, OK, I guess that makes sense. Apr 02 00:24:20 My favorite, though, was the time some people lectured me on how I shouldn't be doing string handling in C, I should be writing in Python. Apr 02 00:24:21 recruiting is hard Apr 02 00:24:32 And I said that, for the app in question, I didn't think Python would be a good fit, but they assured me Python would be fine for it. Apr 02 00:24:41 Yeah. Apr 02 00:27:08 I know people hate them, but the "off the wall" interview questions really are good at getting context-free problem solving Apr 02 00:27:41 I wish there was something that could provide that sort of thing that, you know, actually applies to Software Apr 02 00:28:18 I was super impressed by Triplebyte's interview process, honestly. Doesn't really check for context-free problem solving, exactly, but it does get over a lot of ground. Apr 02 00:28:47 My observation is that I'm a very good debugger. But more interestingly/usefully, I'm *especially* good at "things other people get stuck on". Apr 02 00:29:00 With the counterpoint being that other people are likely to be good at things I get stuck on. Apr 02 00:29:35 But that means that, in a collaborative environment, I'm more useful than my "raw" ability would indicate, as long as other people have different... whatever the term should be. Polarity or something. Apr 02 00:33:58 It takes a team. Sometimes my best coworkers almost *never* agreed with me... made me had to be *really* sure I was right :) Apr 02 00:34:53 s/Sometimes/Some of/ **** ENDING LOGGING AT Mon Apr 02 03:00:02 2018