| | <<<Back 1 day (to 2017/01/24) | 20170125 |
| tor8 | sh4rm4^bnc: I use the Cybook Muse Frontlight (locally branded version). It's got the best software (not saying much, sadly), a great screen, and actual page flipping buttons. Crap build quality though, really flimsy plastic, but it's cheap and I like it better than the Amazon Kindles. | 10:18.31 |
| | And hey, it lets me upload and choose a font of my own choosing, unlike the Kindle. | 10:19.16 |
| irix_user | hey all :) Am trying to compile MuPDF on the venerable IRIX operating system. Am running into some issues. Is this the place to ask for assistance? | 14:04.51 |
| camelopard | irix_user: Yes, it is. | 14:24.52 |
| irix_user | cool. well, first off cmapdump gives me a Bus Error | 14:26.59 |
| | I trawled through the Makefile and did a few experiements on the CLI and yes, it's cmapdump. I then ran it through the native dbx debugger | 14:27.57 |
| | which points me in the direction of fz_try and then on to the macro #define fz_setjmp(BUF) sigsetjmp(BUF, 0) | 14:29.16 |
| | in ./fitz/system.h | 14:29.32 |
| chrisl | You may well find that mupdf isn't terribly strict about aligned memory accesses | 14:29.55 |
| irix_user | so while i can sort of follow the trail, i;m not enough of a c-programmer to really understand what is going wrong... | 14:30.26 |
| | i also get a large amount of this: | 14:31.03 |
| | include/mupdf/fitz/stream.h:290:2: warning: implicit declaration of function 'sigsetjmp' [-Wimplicit-function-declaration] | 14:31.06 |
| | in order to make my life a little easier... i;m using gcc 4.7.1 instead of the native MIPSpro compilers | 14:31.45 |
| camelopard | irix_user: somebody has to debug this on your box. I doubt that one can find IRIX anywhere else. | 14:37.11 |
| tor8 | irix_user: try changing sigsetjmp to setjmp | 14:37.39 |
| irix_user | okay let me try setting to setjmp, then we can move on to seeing how to setup access to my box. | 14:38.48 |
| | brb | 14:39.09 |
| | compiler is running... | 14:44.47 |
| | hmmm... seems to be moving on better... with the setjpm change. now bumped into something else. investigating. | 14:58.06 |
| | a make nuke and restart of compile fixed the new bumps. compiler is chewing through harfbuzz at the moment. lots of warnings. no errors as of yet. keeping fingers crossed. | 15:16.43 |
| | More compile restarts, but less and less errors. At a whopping 800Mhz this stuff goes a bit slower than on modern machines :) | 15:37.22 |
| chrisl | Blimey, I think my last Irix box was 300Mhz..... | 15:39.41 |
| irix_user | yeah - i got one of the later Fuel machines. It's basically a v slow PeeCee :) also have a 200Mhz O2 lying around someplace, but as cool as that machine is, 200Mhz is really too painful in 2017. | 15:41.36 |
| chrisl | I had an Indy, but the psu died. Then I had an O2, but that died, too | 15:43.03 |
| irix_user | hehe - PSU in my Fuel has died twice. managed to find some cheap fuels with low spec to cannibalise. Silly PSU's in Fuel are WTX not ATX | 15:45.42 |
| | Guess it's a bit daft to have emotional attachment to these machines. But it's nice to boot them every so often, use them for a pet project and then let the cobwebs grow back. Makes a change from the vanilla PC mono culture. | 15:47.25 |
| chrisl | irix_user: I can't really criticize - I have a couple of decade or so old SPARC machines here. If they weren't so damned noisy, I'd probably do more with them! Although, I am finding connections to peripherals and displays is starting become a challenge | 15:52.03 |
| irix_user | yes. the noise is a real 'thing'. I've done some fan replacement surgery on the fuel, which has reduced the soundlevels quite a bit. But the Fuel has some environmental monitoring going on and will refure to boot the system if fans arent running at certain (relatively high) RPMs. | 15:55.12 |
| chrisl | The Sun Blade1000 has the same type of monitoring. It also means it can only really run Solaris, as only Solaris has the drivers to control the fans, and slow them down when not required | 15:57.10 |
| irix_user | but yeah - peripherals.... IRIX has an incomplete USB implementation. Basically keyboards and mice. | 15:58.53 |
| | Display is less of a problem on IRIX and on the Fuel. It's DVI out and I can run 1920x1080 on a modern LCD panel. no issues with 13w3 and Sync-on-Green etc. | 16:01.27 |
| chrisl | AH, both my Blade and Ultra60 are VGA display connectors - those are pretty hard to find these days! | 16:05.56 |
| irix_user | yeah. my 2 main LCD panels still have them - but the touchsensitive one i bought a year ago - that's all HDMI. | 16:09.21 |
| Robin_Watts | chrisl: One of my DELLs has VGA, the other 2 don't. That dell is 350 quid rather than 200 though. | 16:10.01 |
| chrisl | Robin_Watts: Yeh, my second monitor still has VGA, but the main one is HDMI and DisplayPort. | 16:10.43 |
| Robin_Watts | yeah, both of which give windows conniptions :( | 16:11.10 |
| | My monitors gone away! Wow! A New monitor! Let me guess what size it should be! | 16:11.45 |
| irix_user | either way, despite all the issues i still really like my IRIX box. It sort of connects me to an era of UNIX(tm) that is all but forgotten today. I also find i have a deeper understanding for what makes unices tick. Eventhough my c-coding chops leave much to be desired :) | 16:11.54 |
| chrisl | I guess if we're really dedicated to the history computer cause, we should be using totally immovable CRTs..... | 16:12.00 |
| kens | Hmm my 2nd monitor has VGA, HDMI and DVI, not sure about the larger one except that it has VGA and DVI | 16:12.09 |
| kens | has an old CRT here too, that uses VGA.... | 16:12.23 |
| Robin_Watts | chrisl: I have one of those on the desk too. | 16:12.40 |
| | I am considering getting shot of it so I have somewhere to put the 3d printer. | 16:12.54 |
| irix_user | heheh - yeah those SONY trinitron CRT tubes were something else. SGI have a widescreen one on the book sit soldl that with turnkey Flame/Inferno solutions to hollywood | 16:13.01 |
| | *had | 16:13.07 |
| chrisl | I binned all of mine - took up *way* too much space | 16:13.12 |
| irix_user | me too | 16:13.17 |
| | all gone | 16:13.19 |
| | ah. | 16:13.26 |
| | bugger | 16:13.29 |
| chrisl | In fact, I think my last CRT was an SGI one | 16:13.38 |
| Robin_Watts | I was hoping that there would be a skip coming when they did the bathroom next week, but... no :( | 16:13.40 |
| irix_user | modern X11 has come to bite me in the arse :( | 16:13.44 |
| kens | Can build without X11 | 16:13.59 |
| irix_user | Xcursor.h and other extensions are now giving fatal errors | 16:14.07 |
| | i know | 16:14.13 |
| | but the whole point of this exercise is to get mupdf to display stuff in X11 and then I'll take that as a departure point to build a new Motif based viewer application for IRIX.\ | 16:14.55 |
| kens | Ummm,..... | 16:15.11 |
| irix_user | I wrote a bunch of UI hacks for xpdf - made it sgi look and feel and gave it Drag and Drop from the IRIX desktop | 16:15.36 |
| | but xpdf is a pain | 16:15.47 |
| | seems far to complicated | 16:15.56 |
| | mupdf seems much more lightweight. which on mhz limited machines is a BIG deal | 16:16.16 |
| | i also wrote a makeshift CBZ/CBR viewer for IRIX.... and guess what mupdf seems to do that too! so i figured i'd give it a shot and see if i could use it as my display/render engine and roll two of my programmes into one. | 16:17.21 |
| | drum roll.... | 16:19.40 |
| | i;m happy to report mupdf-x11 seems to be working | 16:20.02 |
| | :) | 16:20.06 |
| | will test on more CBZ and PDFs | 16:20.20 |
| | give me a few more minutes | 16:20.26 |
| | well, hats off to all of you. it works and it works well! | 16:23.37 |
| | really fat too. | 16:23.44 |
| | *fast | 16:23.47 |
| | compared to xpdf is feels much faster | 16:24.02 |
| kens | Nice to know it can be cajoled into working | 16:25.24 |
| chrisl | irix_user: you may find xpdf has more features in the viewer. It won't have much (if anything) that the mupdf library can't do, but the mupdf viewer doesn't support everything the core library handles | 16:25.43 |
| irix_user | the Xcursor issues are part of the opengl-based viewer. but i;m aurrently only interested in the x11 bits so eventhough the compile doesn't run to the end - it runs far enough for my purposes. | 16:26.34 |
| chrisl | There's a make option to not build the gl viewer | 16:27.13 |
| irix_user | @chrisl - yeah - UI stuff is basically the code i do. So i'm going to see if/how i can wrap mupdf in more UI. | 16:27.16 |
| | the IRIX community at forums.nekochan.net is going to be chuffed! | 16:28.13 |
| | btw. if you guys have a thing for the old unix boxen, Nekochan is a good place to hang out. many curmugeonly old men with their workstation of yore | 16:29.12 |
| | *chuckle* | 16:29.25 |
| chrisl | I think I'm the only one, and I haven't really had time for ages. My SPARC machines mostly get used for a bit pre-release Ghostscript testing | 16:30.30 |
| irix_user | is anybody interested in the minor IRIX adjustments I made to the Makerules file? I realise IRIX is dead platform so you might not be interested :) | 16:30.47 |
| | i do want to understand why the setjmp thing was an issue. IRIX does have the POSIX sigsetjmp() fn and if (as it says in system.h) it can give a speed boost - i'd like to use it. these older Mhz limited machines need all the help they can get. | 16:32.59 |
| chrisl | irix_user: You might want to check for appropriate defines to make the prototype available ( _POSIX_C_SOURCE >= 1 || _XOPEN_SOURCE || _POSIX_C_SOURCE ) | 16:38.38 |
| irix_user | ahh - yes. I saw stuff like that in /usr/include/internal/setjmp_core.h I'll muck around see if I can get it to work. Now that I know the code can be made to compile I'll go back and see if I cansmooth over some of the other IRIX-related rough edges | 16:42.09 |
| | is there any point in fidgetting with -O2 -O3 etc. ususlly there isn;t as mots of the optimisation s that really make a difference are in the code itself. and you guys seem to know what you're doing :) | 16:43.07 |
| chrisl | irix_user: it's doubtful you'd see a difference. But then, we've probably not had the exposure to MIPS that we have to other platforms | 16:44.30 |
| irix_user | fair enough. I'll go back and just smooth the rough edges in the Makefiles some more and then start looking into the UI stuff. Many thanks to all of you for your time and help. Much appreciated! | 16:46.38 |
| tkamppeter | Hi, anyone has an idea what to do with http://bugs.ghostscript.com/show_bug.cgi?id=697504 ? | 17:35.55 |
| Robin_Watts | tkamppeter: fix it? :) | 17:36.51 |
| | It's only a couple of hours old so no one has looked at it yet. | 17:37.35 |
| tkamppeter | The user provided a lot of info, including the GS command line and the backtrace of the crash, and he also told that it happend when using a CMYK color space and not an RGB color space. Does someone of you perhaps see some konwn pattern about what is happening here? | 17:44.02 |
| Robin_Watts | He didn't supply an input file though. I am just seeing if I can reproduce it. | 17:44.55 |
| | OK, I get a SEGV too. | 17:45.40 |
| | ok, for me, on windows, it's dying in cupsGlobalLock. | 17:48.30 |
| | Looks like cups is doing a horrible static init of the global lock. | 17:49.05 |
| tkamppeter | What does this mean? | 17:50.44 |
| Robin_Watts | cups has a cups_global_mutex in it. | 17:54.31 |
| | It locks/unlocks it around doing stuff that needs to happen atomically. | 17:54.51 |
| | On pthreads you can safely say pthread_mutex_t cups_global_mutex = PTHREAD_MUTEX_INITIALIZER; | 17:55.41 |
| | No such safe formulation exists for windows. | 17:55.53 |
| | so you should be calling InitializeCriticalSection directly. | 17:56.11 |
| | If I nobble that, I no longer get the SEGV. | 17:56.22 |
| | So, the short version is that I can't reproduce his error. | 17:58.34 |
| tkamppeter | The reporter of the bug is not running Windows. | 18:01.50 |
| | What means "If I nobble that"? | 18:03.15 |
| Robin_Watts | I am trying to build it on linux to test now. | 18:03.25 |
| | tkamppeter: "if I nobble that" means "if I hackily remove the mutex just for the purposes of testing" | 18:03.56 |
| | Still works for me. | 18:05.35 |
| | If you want to come back to me with a command line and an input file that you know crashes on linux, we'll look. | 18:06.03 |
| | But at the moment I'm going to close it as WORKSFORME | 18:06.17 |
| tkamppeter | OK. | 18:27.09 |
| | And thanks for your help. | 18:30.19 |
| Robin_Watts | tkamppeter: No problem. If you can get us an invocation that goes wrong, we'll certainly look. | 18:33.25 |
| tkamppeter | Robin_Watts, reproducer for bug http://bugs.ghostscript.com/show_bug.cgi?id=697504 posted by OP. | 20:20.11 |
| | Forward 1 day (to 2017/01/26)>>> | |
Log of #ghostscript at irc.freenode.net.