IRC Logs

Log of #ghostscript at irc.freenode.net.

Search:
 <<<Back 1 day (to 2013/09/12)2013/09/13 
kens chrisl ping08:09.26 
chrisl kens: pong08:09.32 
kens having build troubles o Linux.08:09.42 
chrisl Have you done the autogen.sh?08:10.00 
kens Ah, no I only did ./configure08:10.12 
  One moment08:10.15 
  in ghostpdl and ds ?08:10.48 
  err gs...08:10.57 
chrisl What are you building?08:10.57 
kens Just gs08:11.00 
chrisl Then just in gs08:11.06 
kens ah OK08:11.09 
  It was giving an unsupported platform error in openjepg08:12.54 
  Looks like its OK now, thanks08:12.54 
chrisl That's probably because it's now OpenJPEG 2 in there08:13.24 
kens Yeah its been a while since I last updated this VM08:13.50 
  OK it built fine this time, thanks08:14.00 
chrisl NP08:14.06 
sebras see you guys soon!08:49.59 
Robin_Watts gawd. It looks like it takes and hour and 40 minutes to generate the warnings stuff that Marcos does on each commit.11:46.02 
  And it's just caught up to when chris bumped the version number to 9.1011:46.38 
  This could take a while.11:46.42 
  And protecting the dashboard has not protected the reports :(11:49.45 
deleet Robin_Watts: hey, I am planning on implementing an interface to mupdf on android with fragments sometime next month, do you guys have anything planned yet?11:53.08 
Robin_Watts deleet: Fragments? Not at all.11:53.25 
deleet alright, cool. also, does mupdf have any standard format for saving annotations outside of the file itself? (haven't looked)11:54.09 
kens For now and the logs, does anyone have a copy of Windows server 2003 I could borrow ?11:58.48 
Robin_Watts deleet: No.12:29.18 
  ok, I've updated the dashboard so that mupdf warnings have links too.12:33.26 
  and I've updated it so that the ghostpdl ones offer a full set of links, but we'll need to wait for the warnings to catch up to try them,12:33.56 
kens Sounds good to me12:35.35 
Robin_Watts I've just moved all the cgi-bins called by the dashboard so that they should now be protected by passwords.12:51.17 
  You'll all need to reload the dashboard.12:51.25 
  If any of the links are broken, please let me know.12:51.35 
  Flooding in Denver looks biblical.13:34.18 
kens :-(13:41.09 
Robin_Watts Bah. Why is it that when I find some really dodgy code, and git blame it, it's always mine ?13:43.33 
  oh, it's OK. I was clearly smarter then than I am now.13:44.09 
kens diminishing brain cells with age13:44.36 
Robin_Watts http://marcosw.no-ip.com:8080/artifex/mupdf/455fecd8213201417a75f887349a5c3b75d5f46d/mupdf-scan-build.txt13:45.33 
  ok, I see what it's whining about.13:45.57 
russcriptor_ is there a way to have mudraw autocrop?15:00.13 
Robin_Watts autocrop ?15:04.29 
  unhappy cluster :(15:08.46 
kens cannot connect to the dashboard15:10.02 
  ah finally got in,15:10.15 
Robin_Watts kens: really?15:10.16 
kens took a *long* time15:10.22 
Robin_Watts ah, well that may be an indication that the network issues are real :)15:10.35 
kens Indeed, that's why I mentioned it15:10.47 
Robin_Watts thanks15:11.08 
  paulgardiner: http://git.ghostscript.com/?p=user/robin/mupdf.git;a=commitdiff;h=aefe6a511b3cb901034e5995fd882e334e40f2bb15:12.19 
  paulgardiner: Looking at yours now,15:12.23 
marcosw Robin_Watts: I'm going to reboot casper (you are the only one logged on, other then me).15:13.41 
Robin_Watts marcosw: I think I'm disconnected now.15:14.11 
  marcosw: Thanks for the warning.15:14.15 
  marcosw: How do you do your non-web cluster dashboard thing please?15:14.39 
marcosw ~regression/c15:15.01 
  casper is coming back up :-)16:11.49 
  casper is back up and everything is working but slowly. I presume this because everyone abandoned the aws zone with connectivity problems and the other zones are now swamped with new instances being launched. I'm in the process of copying our instance to a different region (from N. Virginia to Oregon), that will take a while and until it's finished there isn't much else I can do (right now the aws console page is overloaded as well, so I can't even se16:27.43 
chrisl Robin_Watts: any chance of getting one of the "Handel & Companye" CDs tomorrow?16:51.46 
mvrhel_laptop this cie to icc stuff is taking too long. it is starting to feel like a wack the gopher game16:56.13 
ray_laptop henrys: You are back from Chi-town, right ?17:00.10 
  henrys: I hope you aren't in the area affected by all that rain. How is it there?17:00.46 
henrys ray_laptop: I'm on high ground and unaffected but this is really awful...17:01.41 
vtorri hey17:02.14 
  is there someone here who can help me with the libs API ?17:02.28 
  arg17:06.18 
  libgs API17:06.22 
mvrhel_laptop henrys: I just heard about it all on the radio. Sounds really bad17:09.33 
henrys the "creek" that runs through my town Longmont looks like the mississippi… Sabrina put a ton of pics on Facebook, mvrhel_laptop 17:13.49 
mvrhel_laptop I will have a look17:14.01 
Robin_Watts chrisl: Will have a look in my loft :)17:25.32 
ray_laptop vtorri: What do you need to know about the libgs API. I assume you've read the doc/API.htm17:29.39 
vtorri ray_laptop: i have a segfault when i call gsapi_run_string_with_length()17:36.44 
  i have created the instance17:36.55 
  i have set the callbacks17:37.10 
  i have called gsapi_init_with_args with the args17:37.56 
ray_laptop vtorri: a debugger should tell you where you are segfaulting from. Are you on linux or windows ?17:38.18 
vtorri i have set a string buffer to "<< /Orientation %d >> setpagedevice .locksafe" with %d replaced by 017:38.25 
  currently on linux17:38.35 
  but i use a pre-compiled libgs (mageia distro)17:38.51 
  so i'm not sure that gdb or valgrind will help me here17:39.10 
Robin_Watts vtorri: You can try with gdb, and see what you get.17:39.34 
  if you at least build your app with debugging, you should be able to see that the args you pass in are reasonable17:40.01 
vtorri http://codepad.org/O2rL9dIx17:41.21 
  does it help ?17:41.26 
  i can provide the code if needed17:42.00 
  note that i call all the gs API in a thread, and only in that one17:43.06 
ray_laptop vtorri: so the function: etui_ps_display_cb_update is your display handler ? It looks like the segfault in in the 'display' device you've set.17:43.20 
vtorri it's the ast function in the display_callback structure so i think that it is indeed the display handler17:44.37 
  last*17:44.39 
  so maybe the loop which calls memcpy is wrong17:45.42 
  at least i have a starting point where to look at the bug17:46.12 
  thank you17:46.13 
ray_laptop vtorri: display_fill_rectangle fills the memory buffer then calls out to whatever display callback->display_update function you'veset17:46.25 
  vtorri: OK. Good luck17:46.42 
vtorri thanks17:46.50 
ray_laptop vtorri: note that gs_fillpage is the very first thing that is done, to fill the page area (in this case 414 by 711 pixels) with white. Then the "update" is called in case your display device wants to paint the screen with the updated page17:48.41 
vtorri ray_laptop: what is display_fill_rectangle ?17:48.49 
ray_laptop vtorri: that is one of the gs 'device API' procs that paints a rectangle with a color. It is FREQUENTLY used. It paints into it's page buffer which can be read out by the 'get_bits_rectangle' device procedure.17:51.18 
vtorri ok17:51.53 
  in my case i want to fill a piece of memory (the data of my graphic object) with : http://codepad.org/V5mqTU0C17:52.27 
  does the loop seem good ?17:52.36 
  pd->gs.**** is what is returned by libgs17:53.07 
  my data has BGRA as colorspace17:54.06 
ray_laptop vtorri: what's in the 'pd' struct, and how is it set up (things like gs.image and gs.stride) ?17:55.47 
vtorri pd is my big structure17:56.27 
  pd->gs is filled with the display_callback callbacks17:56.49 
  in _cb_presize(), should I malloc the image ?17:57.30 
ray_laptop vtorri: OK. So what is the gs struct ?17:57.40 
vtorri ray_laptop: http://codepad.org/46MAfTBe17:58.13 
  and my callbacks : http://codepad.org/Sdln1o3217:59.27 
  ok, first problem : the stride is not width*4, which i have assumed18:03.23 
  hmmm, i think i begin to understand18:04.40 
  _etui_ps_display_cb_update() displays all the "elements" of the page, right ?18:05.05 
  and not the whole page ?18:05.21 
ray_laptop vtorri: when display_fill_rectangle paints, it does so into its memory device's buffer18:05.28 
  vtorri: the update callback can do whatever you want. See psi/dwmain.c which only actually updates the screen periodically.18:06.35 
vtorri ray_laptop: and the page_callback ?18:07.13 
ray_laptop you probably don't want to update on every fill_rectangle call.18:07.14 
vtorri indeed18:07.27 
  i just want the whole page18:07.36 
  i want a void * image and i want to memcpy(pd->efl.m, image, size);18:08.31 
ray_laptop vtorri: the display_page only gets called when the page is complete (from the display device's output_page proc)18:08.36 
vtorri ok18:09.03 
ray_laptop vtorri: so you need to know where the display device has it's buffer that it paints into, and it's "raster" (the stride)18:09.36 
Robin_Watts vtorri: I suspect that gs won't give you the image in one continuous block as you want. You will need to cope with stride != w*4.18:09.52 
vtorri ok18:10.22 
ray_laptop Robin_Watts: that's what 'bitmap_raster()' is for18:10.34 
Robin_Watts ray_laptop: Indeed.18:10.41 
ray_laptop that defines a memory device's "stride"18:10.48 
vtorri time to go18:11.05 
  thank you very much !18:11.20 
ray_laptop vtorri: the device's get_bits proc will return a line of data from the buffer18:12.40 
  vtorri: dev->proc.get_bits (see base/gxdevcli.h for the info on this proc). You specify the line (y) and it returns a pointer. That way you don't need to know how the data is set up in the memory device's buffer18:14.47 
vtorri ok, i did a final test18:15.33 
  the update callback does nothing18:16.02 
  the page callback fills my piece of data with a loop18:16.23 
  and it seems to work18:16.28 
ray_laptop I have to run an errand now ...18:16.32 
  vtorri: OK, good. Sorry I have to run, but I'll check back later18:16.50 
vtorri so i'm happy18:16.56 
  np18:16.59 
  thank you again18:17.03 
ray_laptop vtorri: great!18:17.04 
  vtorri: you are welcome18:17.12 
mvrhel_laptop and one more cluster push to test cie to icc fun18:18.03 
  henrys: I am surprised you have power. I would have thought they would shut parts of it off18:22.10 
Robin_Watts mvrhel_laptop tempts fate on henrys behalf.18:53.36 
henrys really is amazing what's going on: http://www.denverpost.com/portlet/article/html/imageDisplay.jsp?contentItemRelationshipId=540488719:16.22 
mvrhel_laptop ok. cie stuff done for a while, until I tackle V2 profiles for kens19:16.25 
  now onto 694551 which is PCL, FAPI, Overprint 19:16.40 
  very curious about this one19:16.51 
Robin_Watts henrys: It's Gods judgement on the liberals.19:17.15 
mvrhel_laptop henrys: it sounded like there are quite a few people that are cut oof19:17.19 
  yes, the red parts of CO are saying "I told you so"19:17.36 
  marcosw: so I need to put together a desktop machine for my son. what is the best deal to be found without getting something that sounds like a vacuum cleaner 19:21.28 
  ok fair enough. profile_struct is null...19:27.45 
  sounds like my issue19:27.45 
  this is weird though19:28.47 
  lunch break for a bit19:28.53 
  hmm a few warnings in my last commit. will try to get those taken care of20:16.18 
  ok have a fix for 694551 . henrys: can you look athttp://www.ghostscript.com/~mvrhel/ICC_character.patch and let me know if you are fine with this?20:30.26 
  running cluster test on it now20:36.33 
  bbiaw20:38.33 
marcosw amazon claims they've resolved the problem with the connectivity to the us-east-1a zone, so I'm going to put casper back. It will be down ~30 minutes.21:08.14 
  okay, casper is back up. please let me know via sms or email if something is amiss.21:56.25 
mvrhel_laptop henrys: so cluster issues with the above patch. are you fine if I commit?22:18.04 
  s/so/no/22:18.11 
henrys mvrhel_laptop: I tried to look but casper seas down. Is it up now?22:18.33 
mvrhel_laptop yes22:18.44 
henrys okay so why isn't the change in make_mem_mono_device? It is needed anytime you use it right?22:21.22 
mvrhel_laptop well this is the only call to make_mem_mono_device in the whole code base but yes, I could move it in there in case someone uses it in the future22:22.48 
  should I go ahead and add a return also then on gs_make_mem_mono_device_with_copydevice22:23.17 
  in case the init fails for some reason22:23.31 
  which could happen if there is a path issue if we built without the romfs22:24.10 
  henrys: I am fine making that change22:27.43 
henrys mvrhel_laptop: I'm not feeling on firm ground with this one. Why wouldn't the same problem occur with gs_make_mem_mono_device?22:28.03 
mvrhel_laptop let me look. depends upon how it is used22:29.28 
  it is always possible to make a random device and try to use it without having things properly intialized22:29.46 
  in most cases, the device is set up to have the target device profile be the profile22:30.57 
henrys and why the default profile? Suppose somebody is using a custom gray profile?22:31.18 
mvrhel_laptop set_dev_proc(dev, get_profile, gx_forward_get_profile);22:31.53 
  and a comment before that saying /* Should this be forwarding or monochrome profile? */22:32.09 
  with gs_make_mem_mono_device22:32.32 
  which was my concern with the other case22:32.40 
  If someone is using a custom gray profile, it is the default profile22:32.56 
  henrys: so with that aspect we are correct22:33.03 
  I would argue that gs_make_mem_mono_device is actually wrong in forwarding22:33.13 
  if I am going out to a rgb device there would be some issues22:33.34 
  I wonder who made that comment. let me do blame22:33.52 
  and yes indeed. I made that comment22:35.07 
  henrys: so let me go ahead and set both gs_make_mem_mono_device and the other one to get set up with the gray profile22:35.49 
  question though22:36.00 
  is gs_make_mem_mono_device ever called during clist playback?22:36.18 
henrys yikes performance22:36.20 
  every character cached needs a mono men device and we're going to swap out profiles?22:36.59 
mvrhel_laptop right now we would be simply assigning a pointer22:37.25 
henrys oh okay.22:37.41 
mvrhel_laptop no different than assigning the proc22:37.42 
  the profile is already set in the manager.22:37.58 
  however, that brings me to my clist question22:38.08 
henrys you don't dump say a cache of colors when switching?22:38.22 
  mvrhel_laptop: I think clist must use it.22:39.12 
mvrhel_laptop ok. that presents a problem22:39.22 
  the other function was only called during clist writing22:39.43 
henrys and why do we only see a problem in this very isolated situation?22:39.54 
mvrhel_laptop this call is only used with non-FAPI case. Don't know why with this file22:40.17 
henrys mono memory device are created all the time.22:40.26 
  it just seems something else should break.22:40.40 
mvrhel_laptop and bold_fraction != 022:41.30 
  for what ever that is worth22:41.34 
  I am not a font person as you know22:41.41 
  do you run into alot of bold tt in PCL22:42.26 
henrys I'm sorry I'm just puzzled by this one. It really isn't about fonts - if I create a DeviceGray pattern I make a mono memory device and everything works fine.22:43.18 
  maybe you added the icc init to the pattern code.22:44.55 
mvrhel_laptop ok. but I do think that there is a bug in gs_make_mem_mono_device22:45.00 
  as it is forwarding to the target device.22:45.21 
henrys yes I agree.22:45.23 
mvrhel_laptop let me make that fix, and see what explodes22:45.29 
  as that will def. occur in lots of places22:45.42 
henrys stand back!22:46.05 
mvrhel_laptop one question though22:46.09 
henrys go ahead22:48.36 
  very few bold tt tests22:49.05 
mvrhel_laptop sorry. I am back 22:49.35 
henrys I am only concerned that this pattern of making a mono memory device is very common in the code - I don't care about PCL bold.22:49.53 
mvrhel_laptop I am worried that if gs_make_mem_mono_device is called during clist playback, there is no default gray profile to grab22:50.19 
henrys well it shouldn't fail…. but you know what I mean.22:50.26 
mvrhel_laptop as any profiles that are needed must be packed in the clist22:50.43 
  I don't understand how it can work right now the way that it is22:51.02 
  I am going to go ahead and make the change and see what happens22:51.15 
henrys I don't know if clist would use it if the profile were never set before clist.22:51.53 
mvrhel_laptop what I mean by "I don't understand how it can work right now the way that it is" is that the color mappings should be wrong when going to the mono device with an RGB target device22:51.56 
henrys yes that is what is completely nutty how is this working now?22:52.21 
mvrhel_laptop if I were doing a RGB fill in the mono mem device22:52.23 
  then the color would be mapped to the device RGB profile22:52.34 
  and filled in the mono mem device22:52.41 
  which would seem like an issue22:52.47 
  ok. let me fool with it and set it the way I think it should be 22:53.20 
  and see where that takes me22:53.24 
  oh22:53.47 
  gs_make_mem_mono_device has no access to pgs22:53.59 
  you know though22:55.44 
henrys let me step through in the debugger a more common use of the mono device so I can understand this better.22:56.17 
mvrhel_laptop that is what I was just thinking I need to do22:56.38 
  what is a good file to use?22:56.46 
henrys any character will create a mono device to make the cache bitmap.22:57.43 
mvrhel_laptop ok actually, I am going to go ahead and change gs_make_mem_mono_device_with_copydevice to set it to forwarding and then see what is going on22:58.38 
henrys the default gray profile is subtractive right? It can't be just using the first component of the rob profile.22:58.42 
  can it?22:58.45 
  s/rob/rgb22:58.53 
mvrhel_laptop it is additive22:58.54 
  ok here is a call into gs_make_mem_mono_device22:59.18 
  hmm. other than gs_setoverprint we never seem to call to get the profile23:00.39 
  oh I see23:01.39 
  we always do a call to gx_set_device_color_123:01.59 
  I suspect we are not doing any color management here23:03.07 
  as we are sticking in a Gray color space with no profile23:03.23 
  and likely using the DeviceGray mapping proces23:03.36 
  that makes sense now23:03.39 
  and is probably what we want anyway23:04.12 
  for text23:04.32 
  bitmaps23:04.32 
  hmm henrys: I need to look at this a bit more later. I have to take care of some home issues here now.23:09.21 
  sorry to drag you into the quagmire 23:09.36 
henrys oh no problem.23:09.45 
 Forward 1 day (to 2013/09/14)>>> 
ghostscript.com
Search: