IRC Logs

Log of #ghostscript at

 <<<Back 1 day (to 2013/07/07)2013/07/08 
kens tor8 Robin_Watts_ there's a couple of new bugs for MuPDF on stack overflow. This one might merit an answer:08:36.20 
  This one is on hold, and maybe doesn't but it woul dbe good if you could at least take a quick look in case its something obvious:08:36.20 
Robin_Watts_ kens: I've replied to the first. I am minded to leave the second as closed.09:20.33 
kens Robin_Watts_ : I thought that might be the case09:20.44 
Robin_Watts_ Neither really has enough information to answer them, afaict.09:20.53 
kens Probably true, SO questions rarely do09:21.08 
sebras kens: actually it's the same here on irc, but people sometimes stick around long enough for you (or me) to ask questions that paint the bigger picture.09:25.06 
kens Indeed09:25.13 
kens2 Robin_Watts_ : how do I squash a bunch oof commits from a local branch into my master ?10:42.55 
Robin_Watts_ kens2: change onto your local branch.10:43.19 
  git rebase master10:43.23 
kens2 OK10:43.26 
Robin_Watts_ That will move your current branch up so it's all sat on the end of the master.10:43.44 
kens2 done10:43.49 
Robin_Watts_ All you then need to do is to change 'master' so that instead of pointing to a point back from the tip, that it now points to the tip.10:44.10 
  Do you have any other changes in your tree ?10:44.17 
kens2 Only a stashed interim fix10:44.28 
Robin_Watts_ but it's stashed, so that's not in your tree, great.10:44.41 
  So... git checkout master10:44.49 
kens2 Yes, that much I know :-)10:44.51 
Robin_Watts_ then git reset --hard HASH10:44.59 
kens2 hash is the branch commit ?10:45.16 
Robin_Watts_ where HASH is the hash or the name of the branch that points to the tip.10:45.18 
kens2 Hmm, OK10:45.43 
  seemed to work10:45.48 
  I was exepcting to squash in there somewhere10:46.05 
Robin_Watts_ squashing is where you take a series of commits and compress them into one.10:46.31 
kens2 Ah, I am now 5 commits ahead of origin10:46.38 
Robin_Watts_ Were you wanting to squash all the commits into 1 ?10:46.51 
kens2 Yes10:46.54 
Robin_Watts_ OK, so: git rebase -i HEAD~1010:47.04 
  That will open an editor window with the top 10 commits showing.10:47.16 
  Unless you've set it to something different it will probably be vi. Is that OK ?10:47.46 
kens2 I don't think its vi....10:47.57 
  Ah, its VIM, no better :-(10:48.07 
Robin_Watts_ OK, can you get out of it?10:48.18 
  Esc :q RETURN10:48.24 
kens2 Well I can always quit :-)10:48.26 
Robin_Watts_ ok. do so.10:48.31 
kens2 Hmm succesfully rebased it says10:48.43 
  But I geuss that did nothing10:49.00 
Robin_Watts_ kens; Right. If you made no changes to the file, the "successful rebase" should be a nop.10:49.03 
  export EDITOR=<your favorite editor here>10:49.26 
kens2 Lets try write10:49.36 
Robin_Watts_ Is 'write' on your path?10:49.45 
kens2 Should be10:49.57 
Robin_Watts_ Mine is set to "emacs" and emacs is on my path.10:49.57 
  OK. Obviously you can test whether it's on your path by typing write and hitting return :)10:50.20 
  Assuming it is, and you're happy, try git rebase -i HEAD~10 again.10:50.36 
kens2 Its is and it works. rebase still uses VIM10:50.56 
Robin_Watts_ exit again10:51.24 
  echo $EDITOR10:51.27 
kens2 write10:51.41 
  I'm guessing I should change a config file for permanent effect10:52.15 
  and possibly even temporary effect10:52.23 
Robin_Watts_ git config core.editor="write"10:52.55 
kens2 invalid key: core.editor=write10:53.38 
tor8 git config --global core.editor "c:\program files\windows nt\accessories\write.exe"10:53.51 
Robin_Watts_ wot tor8 said :)10:54.04 
kens2 I'll have to find the right path, just a moment10:54.31 
Robin_Watts_ kens2: which write10:55.35 
kens2 THe one I have here :-)10:55.44 
Robin_Watts_ says /c/Windows/system32/write10:55.48 
  no, which is a command. It's a very useful command that tells you where on the path things are.10:56.02 
kens2 OK soreted that, no errors, rebase still uses VIM10:57.01 
  git config -l shows core.editor correctly set10:58.02 
  ah, core.editor=vi as well :-O10:58.59 
  seems I have 2 entries10:59.17 
Robin_Watts_ kens2: Same command as tor8 gave you, without the --global10:59.55 
kens2 OK11:00.00 
Robin_Watts_ You've obviously got a global and a local setting.11:00.16 
kens2 I was guessing that yes11:00.30 
Robin_Watts_ You could just delete the local one, but I can't remember the command offhand for that :)11:00.34 
kens2 OK 2 settings both the same. I think --unset is what you need11:00.48 
  aha, now it works :-)11:01.08 
Robin_Watts_ So, you should see a list of 10 commits.11:01.26 
kens2 Yes11:01.30 
Robin_Watts_ The first word on each line is 'pick'.11:01.36 
kens2 last 5 are mine11:01.37 
Robin_Watts_ This means that when you exit the editor, git will 'pick' each of those commits. i.e. nothing will change.11:02.01 
kens2 OK11:02.07 
Robin_Watts_ What you want to do is to change 'pick' to 'squash' for the last 4 commits.11:02.17 
kens2 So I can change the last ones to squash ?11:02.18 
Robin_Watts_ git will then squash the last 5 together, and present you with a window to edit the new merged commit message in.11:02.50 
kens2 Fair enough done that. OK so I exit the editor now....11:03.05 
Robin_Watts_ yes11:03.11 
kens2 Well it says it was succesful, let me see11:03.37 
  Ah, but no window to edit the commit11:03.51 
  Nope, no changes11:04.06 
  I suspect I can't use a Windows editor for this11:04.24 
Robin_Watts_ Michael uses notpad211:04.35 
kens2 :-(11:04.42 
Robin_Watts_ So I'm sure it can work.11:04.46 
  Do it again, and look at what the path of the file given is?11:05.03 
  When you saved the file back did it add .txt onto the name or something braindead?11:05.28 
kens2 I don't think that's the problem. As soon as the editor starts its says "succefully rebased"11:05.30 
  And yes I think it does add .txt11:05.42 
Robin_Watts_ kens2: It sounds to me like it's not seeing the file once you save it back.11:05.45 
  so it follows the original instructions (i.e. a file full of picks) and nothing changes.11:06.03 
  So the .txt is the problem.11:06.07 
  Can you persuade it to save without doing that?11:06.21 
kens2 It seems not, I'll kick it again11:06.35 
  But... I do see it exiting the rebase while the editor opens11:07.02 
  Possibly it is releasing the file ?11:07.11 
Robin_Watts_ kens2: Maybe.11:07.21 
  What editor do you use on unix ?11:07.34 
kens2 nano11:07.41 
  Which I don't have in msys11:07.49 
kens2 Notepad works better but screws up the line endings11:08.57 
Robin_Watts_ right. Michael uses notpad2, which I think is better at that.11:09.49 
  but if I was you I'd just download nano for windows - at least we know that will work, and be consistent across platforms.11:10.21 
kens2 I don't actually like nano either....11:10.35 
Robin_Watts_ kens2: I like 'Zap'. I make do with emacs.11:11.03 
kens2 Grr and notepad2 uses the 7zip compressor....11:11.17 
Robin_Watts_ brackets looked interesting last time I looked.11:11.23 
kens2 great zonealarm reckons notepad2 is a malicious program....11:15.55 
kens2 will struggle with vi11:17.25 
  OK should the last commit on origin/mast be included ? It seems to be11:19.53 
Robin_Watts_ kens2: We said rebase -i HEAD~1011:20.19 
  That means it will show the last 10 commits.11:20.25 
kens2 Right, but I mean after doing the edit11:20.34 
  I now get the opportunity for a commit message11:20.43 
Robin_Watts_ kens2: You only changed the last *4* lines to squash, right?11:21.04 
kens2 Robin_Watts_ : 5, because I want the last 5 commits11:21.20 
Robin_Watts_ kens: No.11:21.28 
kens2 But its saying tis a combination of 611:21.30 
Robin_Watts_ You want to squash the last 4 lines into the preceeding commit.11:21.43 
sebras doesn't wordpad work for you?11:21.45 
Robin_Watts_ Hence *4* squashes, not 5.11:21.55 
sebras I seem to recall it doesn't mess up line endings at least.11:22.02 
kens2 sebras : no11:22.04 
Robin_Watts_ kens2: so, we need to start again.11:22.12 
kens2 Roq OK so how can I abort this halfway ?11:22.14 
Robin_Watts_ save the file and exit the editor, the rebase will finish.11:22.32 
kens2 ah OK11:22.37 
Robin_Watts_ Then rerun the "git reset --hard" command you did before11:22.53 
  That will put the state back to before the rebase.11:23.01 
kens2 Hmmm OK I htink11:23.42 
  one second11:23.46 
  Yes, thats fine11:23.56 
Robin_Watts_ Now git --rebase HEAD~10 again11:24.06 
kens2 Yeah, will be a minute11:24.18 
kens2 remembers -i11:24.54 
Robin_Watts_ sorry, yes.11:25.01 
kens2 OK last 4 edited, the initial one left alone11:25.44 
  That looks better11:26.22 
  well git log looks a bit confusing, but I expect it'll look better after I push the change :-)11:33.36 
sebras kens2: git commit --amend and fix the message.11:33.50 
kens2 sebras, the message is OK11:33.58 
  its the confusion of branches and squashes11:34.09 
  lines all over the place11:34.14 
Robin_Watts_ kens; Well, you should see master pointing to the correct history.11:34.24 
kens2 Robin_Watts_ : it looks ok (I think), just a little confusing11:34.46 
Robin_Watts_ And your original branch name pointing to the unsquashed version.11:34.47 
sebras Robin_Watts_: kens2: btw, you guys have noticed the "fixup" alternative to "squash" in git rebase -i? that's comes really handy when you only want to keep the first commit's commit message.11:35.16 
Robin_Watts_ sebras: Yes, I use that all the time, but I don't jump right into to explaining it to people.11:35.48 
kens2 sebras in this case the commit messages weren't useful, I didn't want to wast time writing messgaes if the code didn't work11:35.50 
  and the remaining 4 messages were of the 'lets try this and see what happens' sort11:36.15 
  I wouldn't have done a commit at all, but it semed to be the only way I could do a cluster test with the additional (PostScript) file11:36.44 
sebras Robin_Watts_: ah. it took me a looong time to notice it. :)11:37.17 
  kens2: ok, nevermind then. :)11:37.42 
kens2 and now... lunch :-)11:38.12 
Robin_Watts_ tor8: So, looking forwards...11:46.37 
  Are you happy for us to introduce curl as another thirdparty dependency?11:46.54 
  It seems the easiest way to get a cross platform http fetcher.11:47.20 
tor8 Robin_Watts_: it would also help for form submission, I reckon11:49.15 
Robin_Watts_ tor8: ooh, good point.11:49.28 
tor8 and it seems to be the most standard library around for it, but is it any easier than just naively implementing http over a socket?11:49.55 
Robin_Watts_ tor8: Yes.11:50.11 
tor8 I mean, there's a lot of stuff in libcurl that we probably don't ever want to use11:50.12 
sebras tor8: if we have curl then we coudl also choose not to use $BROWSER in x11...11:50.24 
tor8 sebras: you mean for http-linked pdf files?11:50.44 
sebras tor8: exactly.11:50.50 
tor8 and I haven't looked at the api but if it makes it trivial to write stuff I'm not against having it11:50.53 
  optional or mandatory though?11:50.59 
Robin_Watts_ tor8: curl has more than we need, but not stupidly much more, AIUI.11:51.20 
sebras tor8: we could use curl to do an HTTP HEAD and check Content-type. if it is pdf then we download into a memory buffer and render.11:51.27 
  tor8: optional.11:51.33 
tor8 adding more and more dependencies does get us deeper into security vulnerability territory11:51.42 
Robin_Watts_ and it keeps our platform independence from threading crap, pretty much.11:51.44 
tor8 Robin_Watts_: how is building libcurl? gobs and megabytes of source with a bunch of configure crap or a trivial make && install?11:52.34 
Robin_Watts_ tor8: On windows, it's a simple new project.11:52.55 
  (i.e. a new thing alongside libthirdparty etc)11:53.18 
  was dead easy.11:53.24 
tor8 I'm thinking whether to bring it in as we did with openssl or as we've done with previous (smaller, saner) thirdparty libraries11:53.38 
sebras tor8: if you want to pick and choose what files to compile like you do for jpeg, then it is more difficult.11:53.41 
Robin_Watts_ I'll try linux in a bit, but I don't expect it to be configure and crap.11:53.42 
tor8 sebras: I don't necessarily care about picking and choosing files, but avoiding their configure stuff is necessary for cross-platform builds11:54.23 
Robin_Watts_ tor8: There are optional things like SSL support for curl, but we can skip that.11:54.36 
sebras Robin_Watts_: isn't that used for https?11:54.48 
Robin_Watts_ and people can add that back if they want.11:54.50 
  sebras: Right. And we won't support https.11:55.00 
sebras Robin_Watts_: ok.11:55.17 
tor8 both the openssl and curl stuff *could* be done with pipe() / fork() / exec()11:55.34 
Robin_Watts_ Given that we may have openssl in there anyway for the digitial signature stuff, maybe we can add that back in.11:55.35 
  tor8: not on windows.11:55.49 
  But the curl dependency is on pdfapp, currently, not on mupdf itself.11:56.34 
  or rather it's pdfapp that depends on curl, not mupdf depending on curl.11:56.49 
tor8 Robin_Watts_: I would expect a fz_open_http() stream type11:57.01 
Robin_Watts_ tor8: Indeed. But that's a pdfapp provided class currently.11:57.20 
tor8 right. but if we pull in curl as a thirdparty dependency, it should move into the core, IMO11:57.44 
Robin_Watts_ tor8: Currently we have libmupdf and libmupdf-v8.11:58.17 
tor8 if desired, we could make it optional much like openssl. I'll have to try bashing on the curl build system before I decide.11:58.24 
Robin_Watts_ Do we want to have libmupdf, libmupdf-curl, libmupdf-v8, libmupdf-curl-v8 ?11:58.43 
tor8 but going with too many optional features is not a direction I like11:58.48 
  Robin_Watts_: we don't have libmupdf-openssl and libmupdf-openssl-v8 though11:59.16 
sebras tor8: so openssl is required..?11:59.32 
Robin_Watts_ Personally, I liked having it external to the core lib, as it emphasises the fact that integrators can hook up their own http fetchers (and stream implementations) without needing to change the code.11:59.50 
tor8 sebras: it's optional, but either you have it or you don't ... you don't get two separate libs 11:59.53 
Robin_Watts_ s/code/core/11:59.53 
tor8 Robin_Watts_: right. so there's another option -- make it a thirdparty lib for the x11/win32 viewers only12:00.33 
  and use a system provided one. much like we do for libX1112:00.49 
Robin_Watts_ tor8: Right, that's actually the situation I have here.12:01.18 
tor8 I'm happy to go along with that for now12:01.34 
Robin_Watts_ So it's not a core dependency, it's an app dependency.12:01.49 
tor8 it does mean an explosion of win32 project targets doesn't it?12:01.59 
  if it's optional12:02.07 
Robin_Watts_ tor8: No, I have a mupdf-curl target.12:02.17 
  it's not in the configuration list.12:02.23 
  akin to mupdf-v812:02.39 
tor8 mupdf-curl and mupdf-curl-v8 both?12:02.41 
Robin_Watts_ currently I don't have a mupdf-curl-v8, but I could.12:02.50 
tor8 right. that's what I mean by having an explosion of targets (not configurations)12:03.06 
Robin_Watts_ Right.12:03.12 
tor8 mupdf-bare and mupdf-bells-and-whistles targets for win3212:03.43 
  either you have only the source as checked out of git, or all the thirdparty binaries unzipped12:04.08 
Robin_Watts_ tor8: I could cope with that.12:04.21 
tor8 v8 and openssl and curl (and eventually freeglut)12:04.22 
  to keep down the number of projects that need maintaining. I'll leave the decision up to you and paul since you're the ones who have to maintain the project files :)12:04.53 
Robin_Watts_ For things which are easy to just download into thirdparty, then there should be no problems with making them required for a bells and whistles build.12:05.42 
  I mean, disc space is cheap, right?12:05.50 
  rebuilding v8 is a bit of a git, but we provide built versions.12:06.28 
  rebuilding openssl might be tricky, so again, we should provide build versions.12:06.51 
tor8 if building curl from source, we can drop it in as a regular git submodule12:07.22 
Robin_Watts_ curl seems to more in line with the other thirdparty libs (like jpeg etc)12:07.27 
  tor8: Right. That's what I'll try after lunch.12:07.36 
  along with making linux builds work with it.12:07.56 
henrys so if you are going to use curl miles has to update the contract right… provided that the above copyright notice and this permission notice appear in all copies…. same as the other libraries.13:06.32 
Robin_Watts_ henrys: It's only an issue if people pull in curl to their apps.13:07.35 
henrys haven't looked at other libraries.13:07.48 
Robin_Watts_ It's not a dependency for the core.13:07.54 
  But if you want to use the example http fetcher, then either you need to pull in curl, or you need to rewrite the fetching bits.13:08.31 
henrys oh okay I thought we'd just have it in there and everything would just work.13:09.16 
  but if it is as you say we'll leave the contract alone.13:10.05 
Robin_Watts_ henrys: We were discussing whether to have it as a core dependency or not this morning.13:10.35 
  I think the consensus is to have it as an app dependency for now.13:10.49 
henrys argh too many fuzzing bugs … oh well14:46.24 
molqr is there a way to select text in mupdf?15:09.02 
Robin_Watts_ molqr: Yes.15:10.21 
  What platform?15:10.28 
  Right hand mouse drag, I think.15:10.35 
molqr Robin_Watts_: thanks. linux. 15:10.44 
sebras moqlr: and if you want the text to end up in the clipboard then CTRL-c ought to work as expected after highlighting.15:52.46 
kens goodnight all16:21.29 
henrys robin_watts_mac:travelling? 16:46.35 
  marcosw: so the plotter printed it on separate sheets?17:08.03 
Robin_Watts_ henrys: no, just turned the mac on :)17:34.47 
  tor8: ping17:38.16 
  Is there an easy way to do an alert box in x11 I wonder.17:38.53 
sebras Robin_Watts_: we have the searchbox.17:45.58 
  Robin_Watts_: I think currently we just print to stderr.17:46.41 
Robin_Watts_ we do.17:46.47 
sebras Robin_Watts_: e.g. when the search phrase can not be found.17:46.50 
Robin_Watts_ the searchbox may be an idea.17:46.56 
sebras Robin_Watts_: well, something like it perhaps.17:47.04 
  Robin_Watts_: in anticipation of the GTK-viewer. ;)17:47.15 
Robin_Watts_ sebras: I have progressive stuff working on X11 too, but on windows I pop up an alert box to say "this page is incomplete". On X11, it just retries instantly.17:48.41 
sebras Robin_Watts_: I think just retrying is good enough for X11 OTOH.17:49.55 
Robin_Watts_ sebras: Actually, just retrying isn't correct.17:50.20 
sebras Robin_Watts_: no?17:50.29 
  Robin_Watts_: if the page data is never delivered I agree with you.17:50.50 
Robin_Watts_ If we try to open the file, it'll retry repeatedly until the first page arrives.17:50.58 
sebras but the normal case surely must be that all data arrives eventually.17:51.02 
Robin_Watts_ but then if go to page 2, and it can't load it, we get a crap rendering on the screen, and no alert.17:51.22 
  In order to retry we have to go back a page/forward a page.17:51.37 
  on windows we get an alert saying "Page rendering incomplete"17:51.51 
  on linux we get nothing.17:51.55 
sebras Robin_Watts_: don't we paint the window gray before rendering the page?17:52.23 
  Robin_Watts_: and is the assumption that the file is corrupt or is this for a valid file?17:52.40 
Robin_Watts_ sebras: We do, but if it can't get the page data, the page comes out the wrong size.17:52.45 
  sebras: The app knows whether the error is because of a TRYLATER or not.17:53.07 
sebras Robin_Watts_: right and if it is because of !TRYLATER then we handle it the same way as we used to.17:53.44 
  and if it is TRYLATER, we just retry immediately.17:53.52 
Robin_Watts_ sebras: If it's because of TRYLATER, I pop up an alert window on windows.17:54.09 
  but that's just a stderr output on linux.17:54.20 
  Is the search window actually just text in the top corner of the normal window on X ?17:54.39 
sebras Robin_Watts_: yeah, and I guess that's ok..?17:54.40 
  Robin_Watts_: it is.17:54.45 
  Robin_Watts_: there's no fancy toolkit.17:54.53 
Robin_Watts_ sebras: I'd really rather have a second window.17:55.02 
sebras Robin_Watts_: hence the lack of decent UTF-8 input e.g.17:55.06 
  Robin_Watts_: why?17:55.11 
  Robin_Watts_: (given the current implementation).17:55.27 
Robin_Watts_ Because it would be the same as windows.17:55.27 
sebras Robin_Watts_: when the GTK-viewer is done I would probably agree with you.17:55.45 
  searching isn't the same on win/linux (or other platforms) right now either.17:56.12 
Robin_Watts_ sebras: Well, in the GTK viewer, I'd hope we could do better - pages that aren't loaded could have placeholders, and pages that are only part rendered could have an hourglass in the top corner or something.17:56.38 
sebras Robin_Watts_: mmm, and in the meantime I'd vote for doing something really simple.17:57.10 
Robin_Watts_ Well, on that basis, I'm there :)17:57.37 
sebras Robin_Watts_: so either stderr (which might be obscured) or possibly a status-like bar at the top of the window similar to the searchbar.17:57.48 
  Robin_Watts_: we have serious issues with form-filling on x11 too...18:00.23 
Robin_Watts_ yeah.18:00.29 
  which could be fixed (or patched at least) as part of the same project.18:00.52 
sebras not only are the inputs done via stdin, but additionally they require input _twice_ for eaach field.18:00.55 
Robin_Watts_ but then the GTK app (peace be upon it) will solve all this.18:01.16 
sebras Robin_Watts_: well, we're betting on it at least. :)18:01.32 
  Robin_Watts_: I have only been slightly annoyed about the form-filling so far. but not enough to fix it. :)18:01.50 
  (too much GStreamer :-( )18:02.06 
Robin_Watts_ sebras: I bet the 'enter it twice' thing is easy to fix.18:02.14 
sebras Robin_Watts_: I though so beforse, but I never found the problem.18:03.56 
Robin_Watts_ Is it not that the first call is on a DOWN and the second is on an UP or something?18:04.20 
sebras Robin_Watts_: might be.18:05.17 
Robin_Watts_ sebras, tor8: So... we essentially have 2 ways of loading files now.18:28.15 
  We can load 'normally' by following all the xrefs etc.18:28.38 
  Or we can load 'linearized' by using the linearized object.18:28.59 
  With pdf_reference17.pdf, the latter gives us a much faster rendition of page 1.18:29.20 
  and then we get all the other pages fairly fast, but with the wrong fonts.18:29.46 
  until a lot further through the file, we get them with the right fonts.18:30.01 
  OR we can ignore the linearization object, and load normally. We have to load 1/4 of the file before we get anything meaningful displayed.18:31.05 
  and other pages come along fairly swiftly after the first.18:31.21 
  Neither of these seems to exactly match what (say) chrome does.18:31.49 
  I'm wondering if we should attempt some sort of hybrid loading scheme.18:32.09 
  Load the first page (and all its fonts) using the linearised object.18:32.23 
  Then load subsequent pages using the normal page tree etc.18:32.50 
sebras Robin_Watts_: I wonder what poppler does.18:40.46 
henrys 600 dpi 12 foot plotter output - whoopee!18:44.57 
Robin_Watts_ henrys: Scan it! :)18:45.24 
  mail it back as a 24bit uncompressed tiff and ask if it's now correct.18:45.46 
henrys that did occur to me.18:46.15 
sebras Robin_Watts_: poppler does appear to parse the hints table..!18:46.45 
Robin_Watts_ sebras: really?18:47.01 
  Well, that's an excuse for me to parse the hints table too.18:47.17 
  I'd have to detect a broken hints table and drop back to living without it though.18:47.36 
sebras Robin_Watts_: yes.18:47.37 
  Robin_Watts_: but when you say we need to load 1/4th of the file, then you are not taking Range requests into account I assume?18:49.41 
Robin_Watts_ sebras: That's using range requests.18:50.03 
sebras Robin_Watts_: oh.18:50.11 
Robin_Watts_ 25% of the pdf_reference17.pdf file is first page, page tree, xrefs.18:50.35 
  So linearised gets us a big boost for the first page.18:50.57 
sebras ok, so we could skip the page tree and xrefs if we read the linearization object.18:51.09 
Robin_Watts_ sebras: Right.18:51.20 
sebras how big are they?18:51.28 
Robin_Watts_ but without the hints object we can't get to subsequent pages.18:51.35 
  xrefs are at least 20% of the file.18:51.42 
  If I can use the hints object, then we can get subsequent pages quickly as I can easily grab the areas of the file that matter.18:52.20 
  but chrome does a respectable job with a broken hints table.18:52.47 
  I've studied the hints stream a fair amount due to producing linearised output from mupdf.18:53.39 
sebras Robin_Watts_: I guess making use of the linearizatin object is a good start.18:55.00 
Robin_Watts_ sebras: Right. That was the first thing I coded for progressive loading.18:55.26 
sebras Robin_Watts_: that at least gives an impression of a quick first page.18:55.29 
  Robin_Watts_: oh, so this is already implemented?18:55.52 
Robin_Watts_ Using the linearisation object to get the first page out, then 'rough' renderings of subsequent pages is already implemented.18:56.24 
  And it works nicely.18:56.33 
sebras ah, so the hints are what remains.18:56.43 
Robin_Watts_ and doesn't rely on byte requests.18:56.44 
sebras well, it doesn't hurt to try to parse it.18:56.54 
  doing the heuristics for detecting bogus ones will probably be difficult.18:57.06 
Robin_Watts_ I now have byte requests working, and hence can work better than you'd hope with non linearised files.18:57.15 
  so the next thing to try is to add hints in and maybe we can get the best of all worlds.18:57.30 
sebras Robin_Watts_: after that it is only U3D remaining...18:59.37 
Robin_Watts_ OCG first :)18:59.56 
tor8 Robin_Watts_: I'd imagine a hybrid linearized-first-page + normal-remaining-pages mode would be ideal20:51.36 
  but won't help much for android startup times if we've saved a location past the first page when reopening files20:52.10 
  also, no, displaying an alert window on x11 is very complicated20:52.35 
  you have to put it in the same event loop as the main window. there's no way to get a simple "modal" dialog box function.20:53.23 
  rewiring it to display a 'broken page' image for a page and periodically retry should be easier20:53.47 
  would be easier, I mean20:53.57 
shelly henrys: are you still on here?21:19.58 
henrys yes I am21:20.12 
shelly good to finally catch you, just wanted to make sure you got my last update about the remaining integration work21:20.56 
henrys I read that sometime back let me refresh my memory21:24.06 
  the varying results suggest indeterms, right?21:26.01 
shelly I thought they were indeterms but now it has settled down, I was thinking of doing multiple runs before the final commit as a sanity check21:27.22 
henrys 694311 - why is that related to openjpeg2 integration? did openjpeg 1 support gcc?21:28.37 
  one dot difference are usually not significant.21:31.50 
shelly yes it did and the code still works but I can not find an official comment on why the option is removed21:32.16 
  barring the dot difference, it now boils down to the 4 files listed21:33.41 
henrys and those are problems with mupdf as well21:37.09 
shelly I believe so but will need to check on my other machine, will drop you a note tomorrow21:39.12 
henrys if you can reproduce it in mupdf simon can have a go at it and then your bounty will clear.21:39.30 
  assuming you don't want to fool with it anymore21:39.48 
  the bounty for the integration that is.21:40.06 
shelly there are a whole stack of fuzzing bugs that I am itching to tackle.....21:40.48 
  I just need some quality debugging time to work on this, my hours are erratic at the moment21:42.14 
henrys is that all you need? I do process a lot of email sorry if I was non attentive to the issues you raised.21:42.34 
shelly I'm done, catch you another time, good night.21:43.12 
 Forward 1 day (to 2013/07/09)>>>