IRC Logs

Log of #ghostscript at irc.freenode.net.

Search:
 <<<Back 1 day (to 2013/01/02)2013/01/03 
marcosw1 emrosenf: the person to ask is Robin_Watts, but he is in the UK and so won't be online for another 6 hours or so.02:38.40 
emrosenf Many thanks02:39.38 
ray_laptop alexcher: you are working late !06:37.08 
  I saw your response to the "qr" bug. I was also somewhat puzzled by this hodgepodge, but I'll post the code needed to overlay these two files, and see if I can get marcos to give me more info or a contact.06:46.35 
Robin_Watts Morning tor810:57.59 
tor8 Robin_Watts: morning10:58.56 
Robin_Watts Updated reviews etc online for you when you get a mo.10:59.09 
sebras tor8: good afternoon!12:56.34 
kens Hi sebras13:00.21 
tor8 Robin_Watts: int fz_stream_meta(fz_stream *stm, int key, int size, void *ptr); ?13:05.14 
Robin_Watts oh, bugger, did that slip through?13:05.35 
  That's part of the pdfwrite stuff.13:05.47 
  Will fix.13:05.52 
tor8 Robin_Watts: the obj_num_list stuff in the colorspace parsing, why not use the pdf_mark_dict functions?13:07.00 
Robin_Watts cos these aren't dictionaries.13:07.33 
tor8 I thought all non-trivial colorspaces (i.e. all except /DeviceRGB etc) were in dictionaries13:08.49 
kens arrays13:09.00 
tor8 kens: oh, right.13:09.15 
Robin_Watts what kens said :)13:09.18 
kens :-)13:09.32 
tor8 s/pdf_mark_dict/pdf_mark_object/ then perhaps?13:09.37 
Robin_Watts We'd have to bloat the object structure for that.13:10.07 
tor8 or pdf_mark_array, given that it's also a container object so could benefit from the same operations13:10.10 
  to prevent recursions etc13:10.24 
  the new bugzilla looks dreadful :(13:10.49 
  the emails are illegible crap, and the web interface has broken layout everywhere13:11.03 
kens I tried the 'classic' scheme and IMO its worse13:11.07 
  the emails should be fixed now13:11.14 
Robin_Watts tot8: fixed commits pushed.13:11.20 
kens tor8 you may need rto set teh preferences to 'text' in Bugzilla for emails13:11.42 
tor8 kens: it's either broken ascii art tables (due to gmail using prop fonts) or terrible too small courier font in the body :(13:12.46 
kens The HTML email (too small courier) looked so bad....13:13.01 
  But marcos changed it to text only13:13.14 
  For me it works the same as before now13:13.29 
tor8 kens: the html email could be made to look better by just nuking the nasty html table borders13:13.33 
  and setting a proper sized font and not monospaced for the body13:13.45 
kens THe teeny tiny Courier body text is terrible13:13.47 
tor8 but I guess that means hacking the bugzilla code13:13.54 
kens Especially since its *smaller* than the boilerplate....13:14.01 
tor8 kens: I can't even read the tiny courier body text on my windows box. the font hinting completely screws up at that size13:14.26 
kens I htought at first it was just Eudora, so I switched it to use the Microsoft viewer. THen put it straight back becaus eit was (unbelievably) worse....13:15.35 
sebras tor8: at least you are now running a relatively modern bugzilla with fewer security holes perhaps. ;)13:15.36 
kens The HTML emails are terrible, I don't know who thought up the layout but trhey should be beaten round the head with a UI design guide13:16.07 
  Body text (the bit you want to read) that's smaller then the boilerplate ?13:16.29 
tor8 kens: very much so!13:16.53 
  but the web interface html looks broken too! http://ghostscript.com/~tor/stuff/bugzilla.png13:17.25 
kens THat used ot happen before too13:17.45 
tor8 kens: oh... never noticed it before though13:19.50 
  Robin_Watts: Bug 693503: Fix overlong (seemingly infinite) loop of warnings. ... that the right bug number?13:20.51 
kens tor8 I have seen it before, but not freqiuently I admit13:22.15 
Robin_Watts tor8: Yes.13:25.36 
tor8 Robin_Watts: I thought it might have been related to 69350613:27.57 
Robin_Watts No, not knowingly so.13:28.30 
  cluster tests just passed for all those current reviews.13:29.09 
  Zeniko just mailed me about the one with obj_list_num in it, and offers an alternative solution13:29.50 
  http://code.google.com/p/sumatrapdf/source/detail?r=7194#13:30.07 
tor8 Robin_Watts: I do worry about aborting the bf_range parse early13:30.11 
Robin_Watts tor8: I'll confess to being slightly at sea there, but it looked OK to me. Can you see a case where we'd go wrong?13:31.01 
tor8 Robin_Watts: I think I'd be happier with a pdf_mark_array function, seeing as we use pdf_mark_dict similarly13:31.16 
Robin_Watts tor8: OK. I'll think on it.13:31.32 
tor8 Robin_Watts: we early out before parsing the array/string that should be mapped13:31.58 
Robin_Watts ok, so we can safely add a marked flag to arrays without bloating pdf_obj's.13:33.29 
tor8 is the error one of producing too many warnings in the map_one_to_many loop?13:33.30 
Robin_Watts tor8: We produce 2 warnings per iteration of that loop./13:33.55 
  hence the warning suppression doesn't kick in.13:34.07 
tor8 oh... which are the two warnings?13:34.30 
Robin_Watts I can't remember.13:34.37 
  I can find you a copy of the file if you want...13:34.51 
  oh, it's on /home/support/693503/*/3192.pdf.SIGSEGV.b0.243813:35.42 
tor8 Permission denied :/13:36.16 
  oh well, I have sudo access so problem solved. just annoying.13:37.55 
  Robin_Watts: two different alternating warnings with that file13:41.03 
  warning: range limits out of range in cmap Adobe-Identity-UCS13:41.20 
  warning: one to many mapping is too large (60); truncating13:41.20 
  then...13:41.23 
  warning: one to many mapping is too large (60); truncating13:41.29 
  warning: cannot map one to many; table is full13:41.30 
  can't help but think that the file is really broken13:41.35 
Robin_Watts tor8: oh yes, it's a broken file alright.13:42.28 
  but an infinite seeming loop is still bad.13:42.43 
Robin_Watts gets lunch.13:42.51 
tor8 Robin_Watts: yes. I'll investigate some more.13:43.23 
  Robin_Watts: oh, I know why it looks so badly broken...13:44.25 
  <0f00> <0fff> <0f00>13:44.34 
  <1000> <10e00>13:44.35 
  <010010e00> <01ff1 <0100113:44.36 
  <02001 <02ff1 <0200113:44.37 
  <03001 <03ff1 <0300113:44.38 
  error: zlib error: invalid distance too far back13:45.15 
  Robin_Watts: maybe we should give up earlier... here's what happens if I fix the warning duplications:13:53.56 
  error: zlib error: invalid distance too far back13:53.57 
  warning: read error; treating as end of file13:53.57 
  warning: range limits out of range in cmap Adobe-Identity-UCS13:53.59 
  warning: ... repeated 1048578 times ...13:54.00 
  error: expected string or endbfrange13:54.00 
marcosw1 tor8: the html bugzilla issue you mentioned earlier doesn't happen under firefox: http://ghostscript.com/~marcos/bugzilla.png14:39.17 
  I'll check under chrome later today.14:39.52 
  there are a lot of configurable options relating to display of bugs, it might be something we can fix.14:40.23 
sebras marcosw1: tor8: it happens both in chromium and firefox depending on zoom-level (ctrl-wheel).14:47.05 
Robin_Watts zooming in the browser breaks *everything*, IME.15:01.01 
sebras Robin_Watts: exactly, I can get chrome to look good in 67% magnification, and I can get chromium/firefox to break if I zoom...15:08.28 
marcosw tor8: I fired up an instance on amazon with the pre-update bugzilla and it renders the html pretty much 100% the same as the new under chrome (i.e. both are formatted badly): http://www.ghostscript.com/~marcos/old_vs_new.png16:26.38 
  I'll leave http://54.235.194.74/bugs.ghostscript.com/ up for a bit, in case people are feeling nostalgic :-)16:28.05 
ray_laptop mvrhel: How are you doing? I hope you are starting to feel better.16:38.35 
  mvrhel: when you do feel better, I'd like to chat with you about a couple of cust 532's issues.16:39.01 
Robin_Watts emrosenf: I have never built gs for iOS, as a library or otherwise.16:46.40 
emrosenf I see. I imagine building it as a library shouldn't be too dissimilar from the OSX Framework target. Do you have any ideas what I would need to do to modify the build scripts?16:48.42 
Robin_Watts emrosenf: I have no idea about the OSX framework target.16:50.04 
  If it was me, I'd work from the standard makefiles and avoid Xcode like the plague.16:50.53 
emrosenf Point taken. Thanks.16:51.58 
  Do you know about the general architecture of the build scripts? If I want to set a bunch of CFLAGS for the entire build process, are they passed between the different makefiles or do they have to be set for each different one?16:53.01 
Robin_Watts emrosenf: chrisl is the expert here, but in general, set CFLAGS (or XCFLAGS) at the top level and they get passed through.16:56.12 
  tor8: New version pushed.16:57.18 
  henrys, tor8, paulgardiner: My memory was that we had vaguely said we'd have a meeting today. I have nothing to report though (other than slowly plodding through the hundreds of failure reports from 395).16:58.17 
henrys Oh I guess I forgot. I don't have anything really17:00.04 
marcosw I restored the previous bugzilla look to not underline hyperlinks (unless you hover over them).17:02.34 
henrys emrosenf:some reason you need gs vs. mupdf which already runs on iOS17:03.07 
  ?17:03.34 
emrosenf henrys: I am playing with PS and PCL 17:03.35 
henrys emrosenf: that's a good reason17:03.53 
emrosenf Well I'll stay on the IRC and let you all know about the script i I ever have any success17:05.26 
  There's no info on Google -- perhaps this hasn't been done before17:05.51 
mvrhel_laptop hi ray_laptop: I am starting to feel a bit better. hopefully I can get a few useful things done today17:10.34 
kens Goodnight all17:23.52 
ray_laptop mvrhel: When I made the change to limit the interpolation level for cust 532, I had the interpolate code emit an image (instead of copy_color) and it was going into a pattern-clist as rectangles, but I tripped over a case where the clist reader was using tdev->color_info.depth that was the wrong value.17:28.36 
  mvrhel: I found that I needed to update the clist device's color_info in c_pdf14trans_clist_read_update, then use the cdev->color_info.depth instead of the tdev->color_info.depth.17:28.37 
  mvrhel: but I can't make the HEAD code fail because it either puts HL images into the pattern-clist, or it puts 'copy_color' (if I force interpolation pre-clist).17:28.39 
  mvrhel: but I don't see how a pattern-clist with a fill_rectangle would ever work with the HEAD code. I'm going to try and put together a test case.17:28.40 
mvrhel_laptop ok. ray_laptop: my head is still fuzzy to be jumping into pattern clists with transparency17:29.42 
  if you have a case that fails though please open a bug for me to look ovr17:30.06 
  over17:30.08 
ray_laptop mvrhel: OK. I don't blame you.17:31.01 
  mvrhel: are you working on the pdf14_copy_color (or was I supposed to have picked that up) ? 17:31.40 
mvrhel_laptop ray_laptop; no I have been working on ETS stuff17:31.57 
ray_laptop mvrhel: cust 532 was asking, so if you are tied up with ETS, I'll dive into that next.17:32.58 
mvrhel_laptop ray_laptop; yes. I want to get this all back to Max. I don't want him waiting on me. 17:33.34 
Robin_Watts tor8: ping17:35.50 
 Forward 1 day (to 2013/01/04)>>> 
ghostscript.com
Search: