| 00:20.27 | Mirell | Heh, yes. |
| 00:20.29 | Mirell | Cat herding. |
| 00:27.05 | g1powermac | forget the cats. . .try herding ten dogs throughout your house. . . |
| 00:31.14 | *** join/#edev landley (~landley@140.242.26.2) |
| 00:44.46 | Mirell | Heh |
| 01:17.30 | *** join/#edev djerome (~djerome@ip68-2-20-108.ph.ph.cox.net) |
| 01:21.45 | *** join/#edev risca (~risca@wnpgmb0903w-ds01-177-34.dynamic.mtsallstream.net) |
| 02:31.04 | *** join/#edev sterNiX (~LessIsMor@unaffiliated/nu253r/x-0655220) |
| 03:05.12 | *** join/#edev sterNiX (~LessIsMor@unaffiliated/nu253r/x-0655220) |
| 03:39.42 | *** join/#edev mbuf (~user@49.248.86.5) |
| 04:09.40 | *** join/#edev kergoth (~kergoth@covenant.kergoth.com) |
| 04:16.05 | *** join/#edev djerome (~djerome@ip68-2-20-108.ph.ph.cox.net) |
| 06:33.31 | *** join/#edev clio (~andrej@85.159.109.222) |
| 07:36.44 | *** join/#edev amarsman (~marsman@90-145-17-249.wxdsl.nl) |
| 08:38.11 | Posterdati | hi |
| 08:59.22 | Posterdati | please I need help debugging a cortex-m3 with linux (lm3s9b96 + gdb). |
| 10:27.59 | *** join/#edev lfelipe (~lfelipe@enlightenment/developer/lfelipe) |
| 10:41.39 | Posterdati | any cortex-m3 geek around? :) |
| 11:03.56 | *** join/#edev florian (~fuchs@port-217-146-132-69.static.qsc.de) |
| 11:03.56 | *** join/#edev florian (~fuchs@Maemo/community/contributor/florian) |
| 11:10.45 | Posterdati | please I need help to debug a crash on a cortex-m3... |
| 11:21.36 | av500 | I need a steak sandwich |
| 11:34.27 | alex4nder | I'd like some breakfast. |
| 12:01.33 | *** join/#edev g1powermac (~g1powerma@rrcs-173-197-150-24.west.biz.rr.com) |
| 12:01.33 | *** join/#edev g1powermac (~g1powerma@unaffiliated/g1powermac) |
| 12:47.00 | *** join/#edev GeorgeJ (~GeorgeJ@86.124.203.104) |
| 12:47.00 | *** join/#edev GeorgeJ (~GeorgeJ@unaffiliated/georgej) |
| 12:50.09 | *** join/#edev gustavoz (~gustavoz@host166.186-109-12.telecom.net.ar) |
| 13:38.36 | Posterdati | I found the problem |
| 13:38.49 | Posterdati | memset causing processor to hang |
| 13:39.01 | Posterdati | maybe is a toolchain problem |
| 14:22.48 | Posterdati | is arm-none-linux-gnueabi a toolchain for bare metal systems? |
| 14:55.04 | *** join/#edev GPSFan (~kenm@64.92.145.112) |
| 14:55.04 | *** mode/#edev [+o GPSFan] by ChanServ |
| 15:22.16 | *** join/#edev Mirell[Alice] (ZvDj3Jwjip@208.100.5.105) |
| 15:22.35 | Mirell[Alice] | Nifty. |
| 15:41.32 | *** join/#edev War2 (~war2@unaffiliated/war2) |
| 15:46.22 | *** join/#edev landley (~landley@140.242.26.2) |
| 16:01.17 | landley | prpplague: How goes CELF? |
| 16:01.31 | landley | Or, you know, anybody who's there... |
| 16:01.59 | landley | did _ask_ weeks ago if anybody was going to CELF and at the time, when travel arrangements could have been made, the answer was no... |
| 16:02.21 | av500 | dave knew weeks ago since he is a speaker |
| 16:02.30 | *** mode/#edev [+o landley] by ChanServ |
| 16:03.02 | landley | shrugs. |
| 16:04.51 | *** join/#edev War2 (~war2@unaffiliated/war2) |
| 16:06.49 | *** join/#edev risca (~risca@wi-secure-8039.cc.umanitoba.ca) |
| 16:14.14 | Mirell | Morning. |
| 16:22.09 | *** join/#edev nwmcsween (adb72733@gateway/web/freenode/ip.173.183.39.51) |
| 16:22.46 | nwmcsween | hi landley |
| 16:23.00 | landley | nwmcsween: hello. |
| 16:23.04 | landley | Mirell: hiya. |
| 16:23.11 | landley | got the darn POST code reasonably complete. |
| 16:23.16 | landley | And there was much rejoicing. |
| 16:23.24 | Mirell | Congrats |
| 16:23.24 | landley | Doing cleanups to check it in now. Binaries delivered to the testing people... |
| 16:23.27 | landley | Thanks. |
| 16:23.33 | nwmcsween | compiling toybox with clang results in a few errors and warnings |
| 16:24.35 | landley | nwmcsween: Good to know. Can you post 'em to the list? |
| 16:24.42 | landley | And how do I setup clang on linux to try this myself? |
| 16:25.01 | landley | is a bit behind on the toybox submissions, but caught up a bit last night/this morning, and has a 3 day weekend coming up... |
| 16:25.21 | landley | nwmcsween: P.S. Thanks for testing/reporting. |
| 16:26.34 | landley | Sigh. One of my pending toybox todo items is taking the toycans.png background image out, because it makes the page hard to read in some browsers. |
| 16:26.46 | landley | is proud of that logo, but doesn't know what to do with it exactly. |
| 16:28.42 | landley | I have the bigger original image (linked from the end of http://landley.net/toybox/about.html ) which I can presumably scale down and stick in the corner or something... |
| 16:29.43 | landley | It's not that it spells toybox in ascii, it's that the bottom layer is all coke zero. Yes, the logo is null terminated. |
| 16:29.54 | landley | (In accordance with the prophecy.) |
| 16:30.08 | landley | (As foretold in legend and song.) |
| 16:30.11 | landley | (As you do.) |
| 16:30.16 | landley | PROUD, I say! |
| 16:31.54 | lfelipe | landley: setting up clang is easy, there is a step-by-step to compile/install from svn at http://clang.llvm.org |
| 16:32.08 | lfelipe | or you can just use if from your distro (ubuntu at least has it on the repositories) |
| 16:32.14 | landley | ifelipe Do they have releases? |
| 16:32.24 | lfelipe | yep |
| 16:32.28 | landley | Ooh, distro version's worth a look... |
| 16:32.39 | lfelipe | but do try to get an uptodate one |
| 16:32.44 | lfelipe | like, ubuntu 11.10 instead of 10.04 |
| 16:33.21 | lfelipe | and if you are gonna try the static analyzer, then I really recommend going with svn |
| 16:34.29 | landley | lfelipe: When the next ubuntu LTS is released, I... may honestly install gentoo on my laptop or something. Not sure yet. |
| 16:34.46 | landley | But until then, I'm sticking with a boring distro on there because I'm busy with other stuff. :( |
| 16:34.57 | landley | I might try the svn version if it's not too fiddly. Lemme see... |
| 16:35.03 | lfelipe | then your best bet is to install from svn, not too fiddly at all |
| 16:35.12 | landley | (svn? Really? Oh well, I knew apple wasn't big into open source...) |
| 16:35.17 | lfelipe | just involves 3 checkouts, but after that is just configure make make install |
| 16:35.34 | lfelipe | it's not exactly an apple project |
| 16:35.46 | lfelipe | but apple does use svn for WebKit as well |
| 16:35.51 | landley | Centralized source control shows the mindset. "We're a newspaper, not a content aggregator. Employee-written articles are the interesting case, externally submitted content is the exception." |
| 16:36.00 | lfelipe | (both have official git mirrors) |
| 16:36.11 | landley | is not a fan of git, but it's better than svn... |
| 16:36.34 | lfelipe | landley: llvm was an university project of lattner at urbana-whatever |
| 16:36.52 | lfelipe | and the university is still heavily involved |
| 16:37.09 | *** join/#edev ElMonkey (~dog@ip-178-202-140-190.unitymediagroup.de) |
| 16:37.32 | landley | Huh, I do have an 11.10 image over here on the other machine, I forgot. |
| 16:37.51 | landley | 178 updates have been selected, 225.4 mb will be downloaded. Yeah, not really using that box much, am I? |
| 16:38.04 | lfelipe | I did clean up the static analyzer warnings from toybox sometime ago, but it was not with allyesconfig |
| 16:38.23 | *** join/#edev T0mW (~tom@static-68-236-213-104.nwrk.east.verizon.net) |
| 16:38.32 | landley | lfelipe: llvm got interesting when Apple threw money at it for 5 years as its way of saying "Hell No" to GPLv3. |
| 16:38.48 | landley | The university project was _nice_, but there were a half-dozen equivalents at the time. |
| 16:39.05 | lfelipe | landley: I'm not saying that Apple did not invest in it, but it is not exactly their project |
| 16:39.06 | landley | lfelipe: allyesconfig isn't very interesting. Defconfig is interesting. |
| 16:39.23 | landley | The "default n" stuff is half-finished, doesn't really work properly yet code. |
| 16:39.24 | lfelipe | though clang (the c-family frontend), lldb (the debugger) and some others were apple internal projects that were later opened |
| 16:39.33 | landley | "I would not want this in my $PATH on a real system" is what "default n" means. |
| 16:39.39 | lfelipe | landley: yep, just saying that I did not clean the other stuff |
| 16:39.58 | lfelipe | also, did not run after the new round of patches, gotta do it sometime |
| 16:39.58 | landley | lfelipe: Apple took an external project and kept it at arm's length while pouring money into it. |
| 16:40.19 | landley | They did this because "Apple's compiler" could not be reasonably expected to displace gcc, and they wanted to kill gcc (at least on MacOS X). |
| 16:41.09 | landley | This isn't dissing Apple by the way: they made intelligent moves based on their motivations. |
| 16:41.26 | lfelipe | yep, and the project grew quite a bit because of that |
| 16:41.32 | lfelipe | so far I haven't seen anyone hurting |
| 16:41.34 | landley | MacOS X has always been exactly open enough to get people to shut up about it. |
| 16:41.42 | lfelipe | even gcc, they finally accepted the need for external plugins |
| 16:42.06 | lfelipe | they wouldn't have done so if not for llvm's competition |
| 16:42.10 | landley | My main objection to llvm is that it's implemented in C++, and that wouldn't prevent me from using it, just from contributing to it. |
| 16:43.05 | lfelipe | it's mostly a sane subset of C++, I was able to do non-trivial changes to it without having my head blown up by it |
| 16:43.10 | landley | It means that a the smallest self-bootstrapping system including llvm unavoidably includes a libstdc++, which isn't _quite_ as bad as sucking in perl but not something I'd want to port to a new processor I'd invented in an FPGA. |
| 16:43.26 | lfelipe | but yes, that is a problem for your usecase :P |
| 16:43.30 | landley | All C++ projects use a subset of C++, the same way all perl projects use a subset of perl. |
| 16:43.39 | lfelipe | thats why I said "sane" |
| 16:43.41 | landley | And as soon as two projects try to share code, the subset is the union of the two... |
| 16:44.03 | lfelipe | their source code is quite nice, do try to take a look around at sometime if you're interested |
| 16:44.13 | ElMonkey | anyone using a blueboard LPC1768-H, and has some trivial sample project to build with gcc? i cant seem to get my makefile and/or linker script right |
| 16:44.18 | landley | I.E. either C++ isn't portable (not even with the same developers working on two different projects), or the "sane subset" argument goes away... |
| 16:44.52 | landley | lfelipe: what I'm interested in is containing the unavoidably circular dependencies in bootstrapping a base system. |
| 16:45.31 | lfelipe | landley: well, the realm of people working on new processors invented in an FPGA is somewhat small |
| 16:45.43 | lfelipe | so, yes, perhaps for that use case llvm is not the best option around |
| 16:45.58 | lfelipe | though I would point you to projects such as ArchC |
| 16:46.05 | landley | You can't get _away_ from the fact that a kernel needs a toolchain to build, a toolchain needs posix-ish command line utilities to build anything, those two need a C library to run, and the C library needs a kernel. |
| 16:46.14 | lfelipe | which can bootstrap an assembler, linker and other stuff for you |
| 16:46.25 | landley | "The circle of booooooooooooot..." (insert musical note symbol) |
| 16:46.27 | lfelipe | no one did the retargetable compiler for it so far, but it was being investigated I believe |
| 16:46.30 | *** join/#edev troth (~troth@nat/hp/x-gmfwczjetqoywczx) |
| 16:46.44 | lfelipe | landley: seriously, take a look at ArchC later :) |
| 16:47.05 | landley | lfelipe: I once came up with a list of over a dozen open source C compilers that couldn't build the linux kernel, uClibc, or busybox. |
| 16:47.36 | lfelipe | landley: ArchC is not a c compiler, it's a architecture description language that boostraps some stuff for you |
| 16:47.51 | lfelipe | it will generate an emulator, linker, assembler and other things |
| 16:47.56 | lfelipe | gdb hooks, etc. |
| 16:48.08 | lfelipe | not sure what kind of stuff you're doing with FPGAs |
| 16:48.08 | landley | Once you've got those three and a toolchain, you can natively bootstrap out from there. Which means supporting new architectures boils down to "get that working" (plus a side team working on stuff like gdb and strace and qemu support), and then build natively in there and fix what breaks because your toolchain is never going to be that good when you start out. |
| 16:48.11 | lfelipe | or if that was only an example |
| 16:48.45 | landley | lfelipe _I'm_ not working with fpgas, Jeri Ellsworth and such are doing stuff with FPGAs. (In their sleep, in her case...) |
| 16:48.50 | nwmcsween | landley: have you looked at musl libc? |
| 16:49.01 | landley | This is the lady who implemented commodore 64 in an FPGA. As a teenager, I believe... |
| 16:49.14 | landley | nwmcsween: yup. Interesting, but LGPL. |
| 16:49.17 | landley | I blogged about it... |
| 16:49.59 | landley | nwmcsween: http://landley.net/notes-2012.html#18-01-2012 |
| 16:50.27 | landley | it's fighting for a share of uClibc's market, which is fighting for a share of glibc's market, which is fighting for a share of Windows' market. Wake me when that amounts to anything. |
| 16:51.13 | landley | I'm interested in "hey, a billion new android deployments over the next 5 years, what happens when that becomes self-hosting and gets hooked up to a USB docking station with a big monitor, keyboard, mouse, speakers, gigE, terabyte hard drive..." |
| 16:52.06 | landley | nwmcsween: an example USB docking station: http://us.toshiba.com/accessory/PA3575U-1PRP/ |
| 16:52.37 | landley | The mainframe was displaced by the microcomputer, because who needs to ship card decks to another building when there's a terminal down the hall? |
| 16:52.45 | landley | s/microcomputer/minicomputer/ |
| 16:52.49 | landley | purl: quiet you. |
| 16:53.18 | landley | The minicomputer was displaced by the microcomputer because who needs the terminal down the hall when you've got your own machine on your desk? |
| 16:53.36 | av500 | sense a series |
| 16:53.38 | av500 | +s |
| 16:54.05 | landley | The smartphone is replacing the minicomputer (rebranded the "personal" computer by IBM's marketing department, which foresaw the rise of internet porn), because who needs the computer on your desk when you've got one in your pocket? |
| 16:54.19 | nwmcsween | have you asked musl about changing to bsd? |
| 16:54.21 | landley | The machine that's closer at hand beats a more powerful machine that's less convenient to use. Always has. |
| 16:54.29 | landley | nwmcsween: I don't think they can. |
| 16:54.58 | nwmcsween | well only '1' commiter for large variances of ego |
| 16:55.05 | landley | It was hard enough for me to do it with toybox, and I only had 5 contributors to track down. |
| 16:55.17 | landley | And I still couldn't get in touch with one of them and had to remove his code. :( |
| 16:55.48 | *** join/#edev War2 (~war2@unaffiliated/war2) |
| 16:55.49 | landley | nwmcsween: if dalias wants to go BSD with Musl I'm all ears. |
| 16:56.18 | landley | Bionic isn't as crappy as toolbox, though. I took a look at it the other week and it's sparse and full of stubs, but there is actually something _there_. (mostly ported from OpenBSD, but eh.) |
| 16:56.40 | landley | Toolbox was a joke. i looked at possibly incorporating some of its' code, but there just wasn't anything worth bothering with... |
| 16:57.09 | landley | Toolbox would be maybe a week from start to finish to reimplement. Bionic would be more than a month. |
| 16:57.30 | landley | (How much more, I haven't read enough of the code to guess. And I admit I suck at time estimates...) |
| 17:06.54 | landley | nwmcsween: do you think Dalias would be interested in relicensing musl? I'm pretty sure he isn't... |
| 17:08.28 | landley | http://lists.busybox.net/pipermail/busybox/2012-February/077352.html |
| 17:09.19 | *** join/#edev sndcrb (~sndcrb@dsl-hkibrasgw2-fef8de00-126.dhcp.inet.fi) |
| 17:09.25 | landley | He believes that busybox's GPL enforcement is "important": http://lists.busybox.net/pipermail/busybox/2012-February/077360.html |
| 17:10.04 | landley | I've come to see the FSF as no different from the MPAA/RIAA in _thinking_ it represents the interests of programmers/artists when in reality it's a dinosaur in complete opposition to them. |
| 17:12.06 | landley | I don't think copyright _or_ patent is the way forward. I think they're both relics of the printing press that need to be swept away by the internet, the way the printing press swept away "holy writ" from monasteries full of scribes and the resulting priests' authority to order executions and lifelong house arrest. |
| 17:13.41 | landley | Yeah, the printing press was an improvement on what came before. And maybe church law was better than anarchic warlordism. (Although really, feudalism just scaled that up a bit if you ask me.) |
| 17:14.06 | landley | But applying copyright to the internet when you copy and publish with a keystroke? It doesn't _work_ that way... |
| 17:14.28 | landley | You might as well pass a law against thinking certain words. |
| 17:15.30 | nwmcsween | so bzero -> memset? a sufficiently useful compiler should optimize the differences away? |
| 17:15.55 | landley | nwmcsween: if the compiler has a built-in memset, maybe. |
| 17:16.02 | landley | If memset is a library function, it literally can't. |
| 17:16.18 | *** join/#edev madsy (~madsy@188.113.88.111) |
| 17:16.18 | *** join/#edev madsy (~madsy@fu/coder/madsy) |
| 17:16.38 | *** join/#edev calculu5 (~calculus@gentoo/user/calculus) |
| 17:16.39 | landley | nwmcsween: show me a compiler that actually optimizes away the bzero/memset differences, and I'll yank bzero. |
| 17:17.36 | landley | Filling memory with zero is actually a special case: the bss starts zeroed, freshly allocated kernel pages start zeroed... |
| 17:18.23 | landley | There used to be a "zero page" in the kernel that was redundantly mapped, until incrementing/decrementing the reference became a hotspot and then the counter wrapped on 32 bit systems with a lot of processes and high memory... :) |
| 17:18.37 | landley | (You do _not_ wanna free the zero page.) |
| 17:19.08 | landley | memset-with-zero is the most common memset. |
| 17:19.18 | landley | That's why bzero happened in the first place... |
| 17:20.13 | landley | It's not a _big_ thing, but it's just irksome that I need to waste an extra 4/8 bytes each instance. if the library didn't already have it, a wrapper to _add_ it would make sense after about the fourth or fifth use.... |
| 17:21.32 | landley | (Three word size arguments for the memset call, plus a fourth word size for the function pointer, plus however much overhead a the function wrapper contains...) |
| 17:23.24 | landley | I'd ballpark the wrapper in the "20 bytes" neighborhood on a 32 bit system. But then there's the question of the compiler possibly creating an inline memset() assembly dump that's actually smaller than "bzero(pos, len)". |
| 17:23.39 | landley | is still waiting for a compiler that doesn't try to store zero as a CONSTANT. Sheesh. Thanks gcc! |
| 17:24.14 | landley | int x = 0 is "xor bx,bx" or similar, not "mov bx, 00000000" |
| 17:24.19 | landley | And yet... |
| 17:24.40 | *** join/#edev risca (~risca@wi-secure-8039.cc.umanitoba.ca) |
| 17:24.42 | landley | And 1 is xor bx,bx; inc bx; still smaller than mov bx, 00000001 |
| 17:24.49 | landley | Especially on 64 bit... |
| 17:25.18 | *** join/#edev cyronin (~tj@users.757.org) |
| 17:25.32 | landley | But then I'm not a compiler optimizer developer... |
| 17:28.49 | nwmcsween | http://paste.pocoo.org/show/552645/ |
| 17:29.19 | nwmcsween | clang -O3 |
| 17:32.45 | nwmcsween | hmm gcc does something different... |
| 17:34.06 | nwmcsween | http://paste.pocoo.org/show/552648/ |
| 17:35.00 | landley | nwmcsween: possibly -Os? |
| 17:35.09 | nwmcsween | no -O3 for both |
| 17:35.19 | nwmcsween | gcc -O3 -S clang -O3 -S |
| 17:36.44 | landley | In your first pastebin, there are two main()s both returning 0 and doing nothing else...? |
| 17:37.39 | nwmcsween | bzero - http://paste.pocoo.org/show/552650/ |
| 17:37.48 | nwmcsween | memset - http://paste.pocoo.org/show/552651/ |
| 17:37.53 | *** join/#edev dijenerate (~dijenerat@173.225.251.175) |
| 17:38.31 | landley | In the second one: yeah, it looks reasonably optimized. That was memset or bzero? |
| 17:39.25 | nwmcsween | both |
| 17:39.30 | nwmcsween | look at .file |
| 17:40.15 | nwmcsween | so leaq and such is faster than xorl? |
| 17:41.45 | landley | Looks nmcsween: dunno? |
| 17:42.14 | landley | Modern processor tend to throw that sort of thing through an instruction translation thing with pipeline reorganization and such. |
| 17:43.17 | landley | The output is some kinda RISC thing. They're _capable_ of winding up equivalent, modulo how much of the original instruction set fits in the cache, which affects how much stress you put on the memory bus to fetch the cache lines containing the code to execute. |
| 17:43.32 | *** join/#edev War2 (~war2@unaffiliated/war2) |
| 17:45.16 | landley | So generally I pay attention to _categories_ of instructions ala boolean and vs modulus, an branches that require speculation or prefetching out of the 3-4 current L1 cache lines that stay hot at any given time. |
| 17:45.32 | landley | But beyond that, "small code fits in cache better, and if it's more expensive now they'll probably fix it in a rev or two"... |
| 17:46.07 | landley | (Locality of reference covers a multitude of sins, but smaller code pushes back against that too...) |
| 17:49.47 | nwmcsween | so bzero -> memset |
| 17:50.04 | nwmcsween | both compilers produce exact output with either memset or bzero |
| 17:50.23 | nwmcsween | at -O3 (probably -O2 as well) |
| 17:52.56 | landley | Ok. |
| 17:54.19 | nwmcsween | free needs sys/sysinfo.h |
| 17:54.36 | nwmcsween | i'll throw some patches together |
| 18:04.15 | landley | nwmcsween: I added sys/sysinfo.h to toys.h, did I forget to check it in? |
| 18:05.19 | nwmcsween | yep |
| 18:06.55 | landley | I should go to lunch somewhere that has internet access, then... |
| 18:19.00 | nwmcsween | strndupa is gnu? needs a portability check |
| 18:20.22 | *** join/#edev War2 (~war2@unaffiliated/war2) |
| 18:22.43 | nwmcsween | toys.h needs time.h for mdev and sort use of strptime |
| 18:48.58 | *** join/#edev dijenerate (~dijenerat@173.225.251.175) |
| 18:51.52 | *** join/#edev nwmcsween (adb72733@gateway/web/freenode/ip.173.183.39.51) |
| 18:52.08 | nwmcsween | clang vs toybox w/ my fixes - http://paste.pocoo.org/show/552684/ |
| 18:52.28 | nwmcsween | mainly using gnu / posix funcs without checks |
| 18:59.48 | landley | nwmcsween: and lo, rsync has sunk an R. |
| 19:00.17 | landley | nwmcsween: uClibc had it, I can add it to lib/lib.c if I need to. |
| 19:00.44 | landley | nwmcsween: it honestly did build locally, I just did a cleanup pass on 2 newly contributed apps and forgot to check in the third file... |
| 19:02.10 | landley | nwmcsween: pretty sure I'm doing a 2>/dev/null when checking for the clone flags, which implies that clang is spitting error messages to stdout instead of stderr, which is Not My Problem (tm). |
| 19:04.06 | nwmcsween | hmm i'll what standards is toybox going to conform to? |
| 19:04.16 | nwmcsween | clang -std=gnu99? |
| 19:05.08 | landley | The "implicit declaration of BLAH" warning should not be followed by a separate line for each "and it takes an argument!!!!!one!" |
| 19:05.44 | landley | And this isn't a compiler difference, this is a libc difference. What libc are you building against? |
| 19:05.55 | nwmcsween | glic |
| 19:05.59 | nwmcsween | glibc |
| 19:06.00 | landley | "Thing isn't in the headers you've #included" is not the _compiler's_ fault... |
| 19:06.20 | landley | Then why hasn't string.h got it? |
| 19:06.43 | landley | strndupa is in string.hy according to the man page... |
| 19:07.03 | landley | I'm not tied to the function, I'm just wondering what's wrong with the compiler that the change is needed. |
| 19:07.12 | ohsix | isn't it wrapped by some features.h thing like GNU_SOURCE |
| 19:07.13 | landley | I expect this sort of stuff building against bionic, but if this _is_ glibc... |
| 19:07.42 | landley | ohsix: I admit the FSF _is_ a xenophobic narcissistic insular bunch of zombies... |
| 19:08.03 | landley | (I forgot to mention paranoid, but this could go on for a while...) |
| 19:08.12 | landley | Still, clang kinda needs to cope with this. |
| 19:08.25 | ohsix | yea, it says what feature test macros in the man page |
| 19:09.26 | landley | http://osdir.com/ml/linux.busybox/2004-02/msg00001.html |
| 19:09.39 | landley | Lying to FSF code is Standard Operating Procedure. |
| 19:09.42 | landley | It can't handle the truth. |
| 19:13.56 | nwmcsween | hopefully musl goes bsd, I think rich may change it |
| 19:14.03 | nwmcsween | seems open to the idea |
| 19:14.37 | nwmcsween | and musl is actually readable unlike glibc |
| 19:14.48 | *** join/#edev dijenerate (~dijenerat@173.225.251.175) |
| 19:16.31 | landley | I'm all for musl technically. I'm just trying to do something at the moment which excludes LGPL code. (I.E. supplement the android userspace and get people to use it.) |
| 19:17.00 | landley | Busybox has been out longer than android has, if they were going to use it they could have at any time, but there isn't even an android installer for it that doesn't require you to root your phone first... |
| 19:19.49 | *** join/#edev nwmcsween_ (~nwmcsween@d173-183-39-51.bchsia.telus.net) |
| 19:22.00 | Posterdati | landley |
| 19:22.04 | Posterdati | hi |
| 19:22.19 | Posterdati | landley: I found the problem, memset crashes my micro :) |
| 19:22.44 | landley | posterdati: um... what? |
| 19:22.57 | Posterdati | landley: FaultISR() |
| 19:22.59 | landley | seems like you'd never be able to _boot_ if that was the case... |
| 19:23.06 | landley | Ah. That thing. |
| 19:23.17 | landley | Yeah, that's a limiting factor, yes. |
| 19:23.21 | *** join/#edev nwmcsween_ (~nwmcsween@d173-183-39-51.bchsia.telus.net) |
| 19:23.23 | landley | Lemme guess: gcc's inline memset? |
| 19:23.31 | Posterdati | landley: using newlib problem IS NOT gone |
| 19:23.42 | Posterdati | landley: no libc memset |
| 19:23.49 | landley | You need some variant of --no-inline-memset |
| 19:23.57 | landley | It's probably producing armv7 code instead of thumb code, or something like that. |
| 19:24.32 | Posterdati | landley: ok, but it fails as it calls the function not in memset |
| 19:25.23 | Posterdati | landley: memset has got pointers as input coming out from a previous lwip mem_malloc |
| 19:29.05 | Posterdati | landley: I was thinking about heap/stck problem |
| 19:29.20 | landley | Posterdati: could be...? |
| 19:29.34 | Posterdati | a bad pointer |
| 19:29.35 | landley | What's the actual instruction that's faulting? Presumably some instruction somewhere is trying to access something it can't, and generating a fault. |
| 19:29.50 | Posterdati | HFSR = 0x400000 |
| 19:30.32 | Posterdati | UFSR = 1 |
| 19:30.41 | Posterdati | other registers are zero |
| 19:33.06 | *** join/#edev risca (~risca@wi-secure-8039.cc.umanitoba.ca) |
| 19:34.23 | *** join/#edev florian (~fuchs@Maemo/community/contributor/florian) |
| 19:50.26 | *** join/#edev nwmcsween_mobi-1 (~nwmcsween@out-ab-159.wireless.telus.com) |
| 20:03.28 | Posterdati | landley: is very strange, fault come out on memset function call not in the memset function itself |
| 20:04.00 | Posterdati | landley: what does it means? |
| 20:08.36 | *** join/#edev SpeedEvil (~user@tor/regular/SpeedEvil) |
| 20:12.22 | *** join/#edev lfelipe (~lfelipe@201.82.19.44) |
| 20:12.30 | *** join/#edev lfelipe (~lfelipe@enlightenment/developer/lfelipe) |
| 20:28.41 | *** join/#edev risca (~risca@wi-secure-8039.cc.umanitoba.ca) |
| 20:32.51 | *** join/#edev lfelipe (~lfelipe@enlightenment/developer/lfelipe) |
| 21:22.44 | landley | There is a down side to having gotten up at 4am to finish thingy before it was due at work. |
| 21:22.48 | landley | I'd put in 8 hours by noon. |
| 21:22.54 | landley | and a couple hours ago, my brain melted. |
| 21:23.01 | landley | is not being very productive right now... |
| 21:32.38 | *** join/#edev bgamari (~ben@physicsnat56.physics.umass.edu) |
| 21:42.26 | *** join/#edev risca (~risca@wi-secure-8039.cc.umanitoba.ca) |
| 21:47.09 | Posterdati | landley: fault address = 0x20004010 in .ld file sram start from 0x20004010 |
| 21:47.18 | landley | Numbers. |
| 21:47.24 | Posterdati | landley: arm is a damned processors |
| 21:47.28 | landley | Is there an instruction at that address? |
| 21:47.54 | landley | And you're not really using an arm, you're using a thumb. |
| 21:48.06 | landley | They're just calling it arm. They added a new instruction set, then removed the old one. |
| 21:48.36 | Posterdati | landley: no instruction there |
| 21:50.48 | Posterdati | in the .ld I put stack and heap |
| 21:52.55 | Posterdati | but it seems that it is useless |
| 21:54.10 | Posterdati | could it be a bad interrupt? |
| 21:54.54 | Posterdati | is stm32 a thumb too? |
| 21:57.31 | Posterdati | shall I put the stack to end of ram or start= |
| 21:57.32 | Posterdati | ? |
| 22:00.28 | *** join/#edev ElMonkey (~dog@ip-88-153-21-152.unitymediagroup.de) |
| 22:02.39 | Posterdati | :( |
| 22:45.57 | *** join/#edev lfelipe (~lfelipe@186.214.55.153) |
| 22:45.58 | *** join/#edev lfelipe (~lfelipe@enlightenment/developer/lfelipe) |
| 23:26.52 | *** join/#edev bsdfox (~h36sa@c-71-197-83-152.hsd1.ca.comcast.net) |
| 23:26.53 | *** join/#edev bsdfox (~h36sa@unaffiliated/bsdfox) |