IRC Logs

Log of #ghostscript at irc.freenode.net.

Search:
 <<<Back 1 day (to 2013/02/24)2013/02/25 
chrisl good morning kens07:58.48 
kens Hi chrisl07:58.55 
chrisl How were the slopes?07:59.06 
kens Excellent :-)07:59.12 
  349 emails to go through now though :-(07:59.39 
chrisl Really? I thought it had been fairly quiet.....07:59.59 
kens Well that's what my inbox says. Of course a number of those will be regression emails08:00.25 
chrisl Oh, right - I automatically filter those08:00.48 
kens coffee....08:50.53 
Robin_Watts I think that bug 693658 should probably be assigned to chrisl/alex/ray/kens rather than me...11:22.53 
kens Seems reasonable11:23.14 
chrisl Depends where the crash(es) happen, doesn't it?11:23.45 
kens Well, I haven't looked at that11:24.02 
kens is still catching up11:24.13 
Robin_Watts chrisl: Someone needs to triage them. Marcos has done the first stage of that.11:24.42 
chrisl Neither have I, given that it was assigned to Robin_Watts, I assumed marcosw had done so based on where the crash is11:24.54 
Robin_Watts chrisl: That's an umbrella bug. There are potentially hundreds of bugs within it.11:25.19 
chrisl <shrug> Probably need to check with marcosw why it was assigned to you, then.....11:26.32 
Robin_Watts cos the support files came in via me, I suspect.11:26.50 
  I have the MuPDF equivalent one assigned to me.11:27.12 
  And the PDF one currently, though that maybe ought to go to alex.11:27.27 
  I can have a go at them all, it's just a question of time.11:27.48 
chrisl That's true for all of us! You can punt it my way if you want11:30.00 
Robin_Watts chrisl: If we share them out, it means that 3 of us will need to ignore the bugs rather than 1 :)11:31.46 
chrisl That was my plan, yes.....11:32.12 
henrys welcome back kens13:57.33 
kens Thanks henrys13:57.40 
henrys Robin_Watts, tor8 (logs), paulgardiner: since you are late anyway do you want to take another week and get the reflow stuff in 1.2 or would that be too ambitious, risky?14:03.09 
  release-wise14:03.21 
  well you aren't late until March14:04.10 
paulgardiner There's a little tidying that might be worth doing in the app, like greying out icons for features that don't work in reflow mode (e.g., search, text select), but nothing that would take as long as a week.14:07.41 
  Robin_Watts would need to comment on the state of the layout-recognition algoritghm, although looks pretty good to me.14:09.08 
henrys funny I started thinking about it due to an error I made - I went on about reflow in the 1.2 artifex newsletter (not released so I can fix it) and then I thought it really would be best to have this in 1.2 given the current state of affairs 14:10.15 
paulgardiner The power of the subconcious :-)14:15.59 
Robin_Watts henrys: I think we should release without it.14:17.22 
  There is no reason we can't do interim releases of the android app.14:17.49 
henrys sigh14:17.49 
Robin_Watts but at the moment, it's got some n^2 algorithms in it, and the error handling will leak if it runs out of memory etc.14:18.31 
  Putting it on the critical path for the release seems like a bad idea.14:18.59 
henrys it just seems much nicer to have "the release" on Google Play one stop shopping ...14:19.02 
  okay I understand14:19.11 
Robin_Watts releases should be stable.14:19.13 
  google has "release+" :)14:19.39 
henrys I don't know if the "app" world is as stable as traditional software markets but ...14:20.42 
paulgardiner Robin_Watts: what about releaeing with an earlier version of the layout recognition algorithm? Ever since you put styles in, we've had something that was well useable14:20.47 
Robin_Watts henrys: Right, so that's why it seems fine (to me) to do builds for the app store at arbitrary points.14:21.32 
  paulgardiner: It's possible that that is in already, actually. Let me check the git hiustory.14:22.11 
  Ok, so there IS simple reflow in the release.14:22.57 
henrys tor8:see the logs14:23.04 
  great so we're good14:24.29 
  now if only we had a release ;-)14:25.07 
Robin_Watts You might want to say that the current release is simple reflow, and more advanced stuff is coming soon.14:25.14 
  tor8: Did you do the rc2?14:25.18 
henrys Robin_Watts:I will and I am going to reference the press release. Miles will have to coordinate that. Are you ready with that?14:26.09 
Robin_Watts http://ghostscript.com/~robin/release.txt is done as far as I am concerned.14:27.25 
henrys very good14:28.16 
Robin_Watts Today I am looking at valgrind bugs, because html was killing me last week.14:40.13 
tor8 I don't think we should put the now reflow text structures in the release14:40.16 
  changing the text extraction API this release, just to change it again in the next one because it isn't stable yet is not playing nice14:40.47 
Robin_Watts tor8: Yeah, all the new structures are out of it at the moment, but the basic reflow mode is there.14:40.52 
  so I think we're good to go. And it means Henry doesn't need to rewrite his newsletter.14:41.15 
tor8 the basic reflow mode I have no qualms about including. it only affects the android app anyway as well14:41.17 
  right. so I'll build the RC2 today (got distracted on friday)14:41.33 
Robin_Watts I found that I just couldn't make <div> and <span> based HTML output work properly.14:42.19 
  until I started adding style="display:table-cell" etc.14:42.41 
  and now it's pretty much OK.14:42.50 
  I conferred with some friends who are very knowledgable in this area, and the consensus is that that's the best I'm gonna get.14:43.17 
  unless we want to write javascript to do the layout. And I don't.14:43.34 
tor8 I suspect you may be trying to preserve too much of the layout...14:45.36 
  and if we want epub output, table (or table-like css) is out14:46.03 
Robin_Watts tor8: Right. But for html output, I want to keep the structure I've worked hard to extract.14:46.51 
tor8 I (naively) thought all you'd need was margin settings... or have you started doing columns in the output?14:47.47 
Robin_Watts columns are needed for things like index pages.14:48.07 
tor8 right, the spaces between spans14:48.37 
  Robin_Watts: how about the two cases when there are one and two spans per line... do you use plain div and dl/dt/dd lists there14:49.41 
  ?14:49.43 
Robin_Watts tor8: I do not.14:50.14 
tor8 I think for more than two columns (not newspaper columns, those should be linearised) having tables is fine14:50.50 
Robin_Watts <dl><dd><dt> tends to split over several lines.14:51.27 
tor8 but for plain text and lists (which dl lists can do all forms, since you have the "bullet" as a separate item) I think we wouldn't want to invoke tables14:51.28 
Robin_Watts i.e rather than:14:51.31 
  *Foo* Bar14:51.44 
  you get:14:51.47 
  *Foo*14:51.59 
  Bar14:52.01 
  I no longer have the bullet as a separate item.14:52.25 
  I spot hanging indents.14:52.37 
  The code to do spotting of list items is currently disabled, as we don't need it to get reasonable results.14:53.07 
tor8 Robin_Watts: hmm. I've seen dl lists without line breaks. maybe they use float:left styles?14:53.20 
  right, with hanging indents we don't need to worry about actual lists14:53.45 
Robin_Watts tor8: pass. I suspect it's possible using scary css.14:54.02 
  but as the page gets narrow, floats break down in nasty ways with overlaps etc.14:54.16 
  That's why I ended up going back to display:table-cell14:54.28 
tor8 right.14:54.43 
henrys 3 degrees F and a foot of snow a good day to stay home.14:54.53 
Robin_Watts crumbs.14:55.16 
  It's cold here, but not that cold.14:55.25 
tor8 that's even colder than here!14:55.48 
henrys we usually get a few days a year like this but mostly it is nice14:57.03 
tor8 we've had alternating days of thawing and -10C14:57.53 
  dreadful... always deadly slippery14:58.05 
henrys wow now the UK has been bumped by Moodys15:07.18 
Robin_Watts yeah.15:10.21 
  henrys: I'm assuming that paint_path in pxl/pxpaint.c is yours?15:15.20 
  It's possible to get through to the final statement in that function (the gs_moveto) without having ever set cursor.15:16.09 
  hence we get valgrind warnings.15:16.14 
henrys okay I believe you ;-)15:16.40 
  if you give me a command line I'll fix it.15:17.25 
  or just the test file you are using15:17.39 
Robin_Watts http://bugs.ghostscript.com/show_bug.cgi?id=69365515:19.03 
  I almost have a fix.15:19.08 
henrys oh I reported that, I should have looked more carefully I just assumed it was a library problem.15:20.48 
Robin_Watts The first one was. The second one isn't.15:24.40 
  hmm. I've modifed the code so that it can never get to the bottom without calling gs_currentpoint. and it's still failing.15:27.16 
  so the current point in the graphics state must be undefined on entry.15:27.29 
chrisl The gs_moveto() should establish a valid current point15:29.08 
henrys Feel free to assign that back to me if it is in XL you are working on higher priority business15:29.33 
Robin_Watts Ah. gs_current_point is reporting gs_error_nocurrentpoint, and we aren't checking the return code.15:30.16 
  Can we ever have a valid path without having a valid point?15:32.24 
kens I don't see how15:32.41 
Robin_Watts kens: In PS I'd agree with you. rules might be different in PCL though, depending on how paths are created?15:33.15 
kens True15:33.23 
chrisl Does a closepath invalidate the current point?15:33.41 
kens I thikn it sets it to the initial point15:33.54 
  PS/PDF of course....15:34.27 
  newpath will begin a path, and have no current point I htink15:34.51 
henrys if you have no current point likely the move to failed and we didn't check the return15:35.23 
Robin_Watts Cluster testing a fix now.15:39.52 
henrys wow pdf browsing in firefox 19 is nice, much better than chrome from what I've looked at so far15:42.20 
  it also has a nice speedup since I last used it, I might switch back.15:43.24 
Robin_Watts pdf.js, innit.15:43.56 
henrys right15:44.39 
  chrome hasn't even put out a 64 bit build on mac os x. They don't seem to have a personnel shortage, WTF?15:48.04 
Robin_Watts henrys: Why would you want a 64 bit build?15:48.42 
  Chrome uses a process per tab model.15:48.58 
  If any one web page needs more than 4Gig, you're doing something wrong :)15:49.19 
henrys I have several plug ins that depend on it - garmin is one. I don't know why.15:51.36 
Robin_Watts possibly because the plugin and browser have to follow the same calling conventions (size of pointers etc)15:57.02 
henrys firefox does seem much more lively to me but it may be because chrome is cookie and crap laden from usage.16:04.57 
chrisl kens: ping16:14.46 
kens pong16:15.02 
chrisl How would you feel about pdfwrite having its own directory in the source tree?16:15.22 
kens fine by me16:15.31 
  Assuming you will update all the builds :-)16:15.45 
chrisl So, I'm thinking we'd have "devices/pdfwrite"16:16.03 
kens Probably ps2write and maybe txtwrite should go in the same place16:16.08 
chrisl ps2write yeh, not really much option there, I can put txtwrite in there, too. I wasn't sure how much shared code there is, there16:17.08 
kens FOr txtwrite, not so much, at least at present, but there is some (the Adobe Glyph List for example)16:17.31 
chrisl There are also some files shared with pswrite, but i wasn't going to bother with that - in the hope it will disappear at some point!16:18.17 
Robin_Watts chrisl: Maybe a dir for vector output devices?16:18.32 
  pdf/ps/xpswrite16:18.41 
kens I certainly hope it will go awqay, just as soon as I can get round to a eps2write16:18.50 
henrys gdevpx, and the tracing device. It would be really nice to do that.16:19.21 
chrisl Okay, I can do that - "devices/vector"16:20.01 
kens makes some sense to em16:20.14 
  me*16:20.18 
chrisl Question then is, would we still want a directory specifically for pdfwrite16:20.46 
kens ps2write and pdfwrite are so toangled up you can't spearate them16:21.05 
chrisl When I say "pdfwrite" I mean both pdfwrite and ps2write16:21.25 
kens Well, I don't know if they need a directory of their own, is this just for neatness sake ?16:21.57 
chrisl Some logical separation - pdfwrite/ps2write is *by far* the largest device, with the most files associated with it16:22.44 
kens Up to you, I don't mind either way16:23.03 
chrisl And what about the PCL and PXL devices?16:23.44 
henrys anything that is a vector subclass right?16:28.23 
  there are no PCL devices I know of just PXL16:28.59 
chrisl the trace device isn't a vector device, is it?16:29.29 
henrys it isn't and if you are going to name in vector and not something like "high level" I'd leave it out.16:30.14 
  I don't know vector seems too developer oriented not user oriented but I don't feel strongly about it.16:31.07 
chrisl This is where the source files are stored - it is *entirely* developer oriented.... no "user" should even be aware of this!16:31.58 
henrys Robin_Watts:can you expand upon non file operations in MuPDF - did you mean stream support or url reading?16:32.08 
  I'm looking at your release notes for 1.216:32.28 
Robin_Watts yes.16:34.08 
  and "yes"16:34.13 
  Essentially you can now open a MuPDF document by passing in a stream, rather than a file name.16:34.38 
  We require the stream to be seekable, currently.16:34.56 
  but it means that people can implement their own 'whole-file' DRM systems (one of our customers does that).16:35.27 
  In future it means that we will be able to open files as they download (I have code that shows this working on branch).16:36.03 
  It also means that people can invoke mupdf on files held in memory without having to go to file.16:36.40 
  henrys: Could you review the patch on here please? http://bugs.ghostscript.com/show_bug.cgi?id=69365516:37.57 
  That's my proposed fix.16:38.03 
  Damn. Wrong file.16:38.20 
henrys wrong patch16:38.36 
  oh you see it.16:38.41 
Robin_Watts Right patch there now.16:39.36 
  Hmm. I sanitised the indentation :(16:39.44 
kens chrisl I fixed that problem with txtwrite, pity the SO user didn't open a bug :-(16:47.32 
chrisl kens: yeh, he wasn't very helpful, really. I got a weird crash when I tried it, though - it crashed accessing SingleGlyphList16:49.22 
kens Yes, the code was barking mad wron g16:49.39 
  I must have been having a bad day when I wrote it16:49.54 
  It seems to work more or less OK now, certainly much better16:50.19 
chrisl I wasn't going to mention it, if the silly sod wouldn't open a bug, then......16:50.50 
kens Well it piqued my interest, and teh crash was a Bad Thing16:51.07 
marcosw henrys: did you see the email from the customer re. the fonts? He does want the 80 fonts from the HP LaserJet set.16:54.31 
henrys so the answer is ghostpdl/urwfonts, no?16:55.58 
  marcosw ^^16:56.08 
marcosw henrys: yes. 16:58.32 
henrys yeah I got one right16:58.48 
  Robin_Watts: sorry distraction reviewing now16:59.24 
Robin_Watts henrys: np. I'm staring at shading code now, trying to understand what !unlinear means.16:59.54 
henrys a no whitespace patch would be easier for me.17:00.54 
Robin_Watts Let me try and make one.17:01.02 
  http://ghostscript.com/~robin/patch.txt17:02.24 
henrys nice thanks17:02.46 
Robin_Watts np.17:02.52 
henrys I should just break down and bring PCL and XL indentation up to speed17:03.26 
Robin_Watts I tried to maintain indentation, until I got to a point where I found that I was getting confusing results, so I reindented in the hopes that it would make it clear why.17:04.17 
  but it didn't.17:04.23 
  I could reapply the fix to a vanilla file if you want.17:04.34 
henrys no I'm good17:04.42 
  Robin_Watts: I'm probably missing something gs_currentpoint returns 0 is there is a path. Aside from that the code could be made much clearer with gx_path_is_null() readability wise, that is what you are after, right?17:11.59 
Robin_Watts henrys: There are 2 separate problems in the file.17:12.29 
  function.17:12.31 
  Firstly, we can get to the final statement without cursor ever having been written.17:13.01 
  and so we can end up setting the current point to an unspecified position.17:13.28 
  I fixed that by adding various gs_currentpoint calls.17:13.42 
  and that lead to the second problem; we were STILL getting undefined values there.17:14.10 
  so I fixed the gs_currentpoint calls to detect whether we had a currentpoint or not.17:14.33 
  If we don't have a currentpoint, then we don't reset the point before we leave.17:14.52 
  So the net effect of this function is that the point doesn't change.17:15.09 
  But I am aware that I may be patching a symptom rather than patching the cause. Hence me asking you to look it over.17:15.47 
henrys right I'd like to dive into it a little deeper by looking in the debugger.17:17.10 
Robin_Watts go for it.17:17.47 
kens OK I think I'm caught up, so I'm heading off. GOodnight all17:18.01 
Robin_Watts night kens17:20.02 
henrys Robin_Watts: more thickness on my part, how does the stream stuff help with DRM?17:20.47 
Robin_Watts henrys: Our customer supplies both reader and content, right?17:21.11 
  They don't use the standard adobe DRM.17:21.38 
  they use their own one.17:21.48 
henrys got it. thank you17:22.04 
Robin_Watts So either they need to decode the entire file, and write it to disc unencrypted, for us to then read back in... 17:22.13 
  np.17:22.21 
  My PC is messing about. better reboot.17:28.39 
henrys if anyone is interested: http://ghostscript.com/~henrys/newstake2.txt -- if you critique it a lot I'll take that as volunteering to do it next time ;-)17:40.50 
Robin_Watts "let's" ?17:42.36 
  Hmm. Professor google says you are right.17:43.42 
henrys I thought let us17:44.07 
Robin_Watts yes indeed.17:44.15 
  I personally find that sentence runs on a bit.17:44.23 
henrys but it is difficult for non english speakers I'll change it.17:44.23 
Robin_Watts I'd prefer a hyphen before the let's, I think.17:44.38 
  or a semicolon or something.17:44.45 
  "All CMYK device" should be "All CMYK devices" ?17:45.24 
  similarly profile -> profiles in the next sentence.17:45.46 
  treatment -> treatments? (arguable, I guess)17:46.16 
henrys argh missed that17:47.37 
Robin_Watts "Further improvements... are include in 9.06".17:47.46 
  I think you need to remove "are included in 9.06" from that sentence.17:47.59 
  "The pdfwrite device now produces output consistent with the linearized PDF specification." <- That's unkind to ken :)17:49.10 
  The pdfwrite device now produces linearized PDFs that work, despite the specification being broken :)17:49.45 
  "MuPDF 1.2 new feature set" => "MuPDF 1.2s new feature set" ?17:50.43 
henrys Ill wait for other to weigh in and fix everything all at once.17:50.44 
Robin_Watts Never start a sentence with But (so I was told at school)17:51.25 
  but ( : ) ) other than those small niggles, it looks great.17:51.58 
henrys actually I didn't really mind writing it … the new stuff in MuPDF is really pretty exciting - Oh god I need a life17:53.00 
chrisl Just a couple of erroneous mentions of 9.06, that I see - starting a sentence with "But" is like using a "goto" in C.... only do so with care!17:53.21 
henrys Yeah I don't know if I'll change the but or let's. BUT I might17:54.31 
Robin_Watts "But for the case of sentences like this, sentences should never start with a but." :)17:57.40 
chrisl I haven't even had anyone bitching at me about using British spelling in the release notes this time - I must be getting better.... or worse?17:59.22 
henrys I put mvrhel_laptop on the hook for some kind of window mobile app by 1.318:04.52 
mvrhel_laptop henrys: that should not be a problem18:12.05 
  I have the zooming, scrolling etc working now18:12.14 
  working on adding search this week18:12.20 
Robin_Watts :points to the "mobile" in that sentence.18:12.25 
henrys Robin_Watts:I'm hesitant to wave the Affero flag in newsletters, can we drop that part?18:12.40 
mvrhel_laptop Robin_Watts: yes. the way I am writing things it should work on the surface and the phone18:12.46 
henrys or press releases18:12.51 
Robin_Watts henrys: If you'd like it dropped, we can drop it. I felt it deserved a mention though, lest people miss it and then complain that we changed silently.18:13.28 
mvrhel_laptop that was odd18:14.06 
henrys well that was my insidious evil plan if you must know - it really is mupdf irrelevant though18:14.19 
Robin_Watts If our intention is to use it to drive people that might have been using MuPDF or gs on servers to having to come to us for a commercial license, then we need to crack the whip otherwise they won't know we're holding one.18:15.02 
  We get occasional questions on here from people wanting to use mudraw to make pdf thumbnails etc. I reckon that could be a web based thing...18:15.55 
henrys we generally know who they are in the past we couldn't do anything about it. Anyone with any success is going to get on our radar18:15.59 
Robin_Watts but yes, less important for mupdf than gs, I will grant you.18:16.19 
  Fair enough.18:16.23 
  I'll happily see it removed.18:16.36 
chrisl It's not relevant at all for a news letter aimed at our commercial customers18:17.47 
henrys right or a press release18:18.10 
mvrhel_laptop need to head out for a bit18:20.57 
ray_laptop mvrhel_laptop: Robin_Watts: I have partially fixed the SMask issue with bug 693115, so that we get the right yellow, red and green colors including the overlap (comparing to Photoshop CS3), which make sense from the RGB values and the smask alpha values for smask118:48.26 
  mvrhel_laptop: Robin_Watts: it also gets rid of the multiple rendering of the SMask that was happening. Now I have to fix the smask2 part (my change is causing it to ignore smask2)18:49.31 
  but it's MUCH better than before, and I didn't have to get rid of the delayed SMask rendering (although I'm not sure if that is ever really of value)18:50.26 
Robin_Watts Will this cause us to get the sane nesting of renderings ?19:24.55 
ray_laptop Robin_Watts: sorry, was debugging. I _hope_ so, but that comes after finding out why the blue is painting as opaque.19:34.51 
Robin_Watts ok.19:35.02 
  My memory of this is superbly rotted now, but I do remember the call sequence as seen from the C side being bonkers.19:35.32 
ray_laptop Robin_Watts: OK. So what is screwy is that the main Contents stream is not in a group, so it paints the Yellow without an SMask (as it is supposed to), but after (now) properly doing the red and green and blending in them using smask1, it finds the inline rectfill of the blue and it paints that into the top buffer, the same as when it did the yellow.19:55.25 
  this seems legal, but it confuses our transparency code. Our code needs a group (that it can pop) in order to paint the blue with the SMask19:57.24 
Robin_Watts I *think* mupdf introduces groups at various points to cope with this kind of thing.20:09.12 
ray_laptop Robin_Watts: probably. So if I change the file to "Do" the blue square following /smask2 gs, then we get EXACTLY the same thing as Photoshop CS3 !20:11.39 
  (with my changes to fix the SMask stuff)20:12.13 
Robin_Watts fab.20:12.14 
ray_laptop so I need to detect the case when a group that didn't used to have an SMask gets one set by and 'gs' operation20:13.03 
Robin_Watts yeah.20:13.38 
ray_laptop and push a group for painting, and pop it sometime later20:13.40 
Robin_Watts "sometime later" = "when the smask is removed from the graphics state"20:14.10 
ray_laptop and the 'sometime' is at the end of the stream, or when when the SMask changes due to a 'gs'20:14.25 
Robin_Watts Due to a new gs, or due to a Q or due to the end of the stream.20:14.59 
ray_laptop need to think about this, but I probably have to hack this into the PDF interpreter 20:15.12 
Robin_Watts (but not the end of ANY stream, the end of the stream in which it was introduced).20:15.30 
ray_laptop Robin_Watts: oh, right -- a Q might pop the old SMask20:15.38 
Robin_Watts Probably easiest to do "new gs, or Q", and then arrange to q/Q around streams or something.20:15.54 
  yeah. nesting hell.20:16.06 
  It looked like PDF interpreter hackery was required to me, which is why it ended up getting foisted on alex.20:16.25 
ray_laptop Robin_Watts: I can (and have) worked on the transparency handling in the PDF interpreter. I did the PS part of many of the bug fixes here (to support mvrhel_laptop)20:17.24 
Robin_Watts oh, yes, I completely believe you're capable of it. I just know that for me, it would have been a massive undertaking.20:18.08 
ray_laptop and probably for mvrhel_laptop as well (since it first requires learning PostScript, then deciphering the mess that is the PDF interpreter)20:19.10 
  mvrhel_laptop has done some reading of the PDF interpreter parts, but wasn't confident on making substantial changes, although iirc, he has done a couple20:20.12 
  anyway, time for a lunch break and let this ruminate...20:23.13 
  bbiaw20:23.16 
 Forward 1 day (to 2013/02/26)>>> 
ghostscript.com
Search: