| | <<<Back 1 day (to 2018/09/24) | 20180925 |
| vtorri | hello | 03:25.39 |
| mubot | Welcome to #mupdf, the channel for MuPDF. If you have a question, please ask it, don't ask to ask it. Do be prepared to wait for a reply as devs will check the logs and reply when they come on line. | 03:25.39 |
| vtorri | in https://bugs.ghostscript.com/show_bug.cgi?id=697122 it is said that we can display pdf with banding | 03:26.08 |
| | is there any code that shows how to do that ? | 03:26.26 |
| sebras | vtorri: it might be easy code, but there is an example in source/tools/mudraw.c, search for fz_new_pnm_band_writer() and the subsequent calls. | 09:15.45 |
| | it might NOT be easy code, it what I meant to say. :) | 09:15.58 |
| tor8 | Robin_Watts_: tor/master has a handful of commits, then I think we should be ready to tag an RC | 12:02.18 |
| Robin_Watts_ | tor8: Looking now. | 12:02.28 |
| tor8 | the try/catch thing probably shouldn't be there | 12:02.49 |
| Robin_Watts_ | So js accepts unix paths even on windows? | 12:06.39 |
| | get_history_filename looks like it will return paths/separated/like/this | 12:07.39 |
| pink_mist | windows accepts unix paths just fine | 12:07.49 |
| | js doesn't need to do anything | 12:07.56 |
| Robin_Watts_ | and even /tmp/.mupdf.history in some cases. | 12:08.03 |
| tor8 | Robin_Watts_: the file reading is done by mupdf not mujs | 12:08.11 |
| | fz_open_file | 12:08.15 |
| | bah, fz_read_file | 12:08.21 |
| Robin_Watts_ | OK, as long as you've tested that on windows, I'm happy. | 12:08.41 |
| tor8 | it seems to work, at least I now have a c:\documents and settings\admin\.mupdf.history file locally on my winxp VM :) | 12:09.42 |
| Robin_Watts_ | tor8: So, the try/catch stuff. | 12:15.14 |
| | It's an extra function call on every always and every catch. | 12:15.27 |
| | but I like the idea of it. | 12:16.00 |
| | oh. and an extra one on every try too. | 12:16.31 |
| tor8 | Yeah. Last I played with this commit I did some measurements and couldn't measure anything, the background noise variation is too large to notice anything. | 12:17.01 |
| | Robin_Watts_: yeah. it does add a lot of function calls. | 12:17.20 |
| | but if it will let us make the fz_context an opaque struct it would be good | 12:17.51 |
| Robin_Watts_ | tor8: yeah. I remember, and I'm grateful for that. | 12:17.55 |
| | I just worry about slowing down something that is so ubiquitous. | 12:18.08 |
| tor8 | Robin_Watts_: considering how inefficiently setjmp is implemented in wasm I'm surprised it even ran at a decent speed! | 12:18.40 |
| | IIRC it trampolines back from the fast wasm bytecode into javascript to do a JS try/catch for the setjmp/longjmp | 12:19.28 |
| Robin_Watts_ | I think I'd argue that we should do the release, THEN put this in. | 12:22.32 |
| tor8 | Robin_Watts_: sure. same thing for the history patch too? | 12:23.01 |
| Robin_Watts_ | It looks like a good idea, but it's not something we should drop in just before a release, given the potential for subtle problems. | 12:23.03 |
| | I don't care about the history patch, cos it's -gl only :) | 12:23.22 |
| tor8 | alright :) | 12:23.29 |
| | updated tor/master with the try/catch after where I intend to stick the RC tag | 12:23.59 |
| Robin_Watts_ | Ok, lgtm. | 12:25.05 |
| Robin_Watts_ | lunches | 12:25.49 |
| vtorri | sebras thank you | 12:36.29 |
| sebras | ugh. I've found a number of leaks today. :( | 13:32.18 |
| Robin_Watts_ | tor8: Urgh. git problems. | 14:32.32 |
| | So, on the windows cluster nodes, we run git under WSL. | 14:32.50 |
| | and the checkouts go onto windows discs. | 14:33.04 |
| | because git is run under linux, we get linux line endings. | 14:33.18 |
| | which upsets stuff when we run it on windows. | 14:33.43 |
| | Ordinarily, I'd set core.autolf = crlf to solve that. | 14:33.54 |
| | BUT the use of .gitatributes * text=auto buggers that up. | 14:34.16 |
| | tor8: https://ghostscript.com/~regression/release__mupdf/ | 14:36.46 |
| | sebras: Did you change colors in BMP? | 14:38.15 |
| sebras | Robin_Watts_: not that I know of. | 14:42.25 |
| tor8 | Robin_Watts_: text=auto should only affect what goes into the git repo (normalize the line endings) | 14:42.48 |
| | Robin_Watts_: is it just the vcproj files that are problematic, or other things too? | 14:43.16 |
| Robin_Watts_ | tor8: It's buggering up submodule updates. | 14:43.50 |
| tor8 | really? that sounds ... insane. | 14:44.27 |
| Robin_Watts_ | tor8: I think it's cos of the aforementioned git:// slowness. | 14:45.10 |
| | I did the checkout on windows, and now the linux side sees every file as being different. | 14:45.26 |
| sebras | Robin_Watts_: what page did you find the bmps on? | 14:45.31 |
| Robin_Watts_ | https://ghostscript.com/~regression/release__mupdf/compare2.html | 14:45.44 |
| tor8 | Robin_Watts_: oh, cause it thinks it has lf endings in its checkout but is seeing crlf? | 14:46.22 |
| Robin_Watts_ | tor8: Yes. | 14:46.31 |
| tor8 | git reset --hard on the linux side? | 14:46.38 |
| Robin_Watts_ | Doesn't want to work. | 14:46.58 |
| | Well, rather, that might work, but it leaves all the files with LF rather than CRLF. | 14:47.15 |
| tor8 | Robin_Watts_: hm, can you rerun that bmpcmp with only pgmraw.72 so we don't have to look through so many duplicates? | 14:47.19 |
| | Robin_Watts_: isn't that what WSL expects? | 14:47.31 |
| Robin_Watts_ | tor8: But it's not what a windows build node expects. | 14:47.49 |
| | Calling VS on project files which are all LF rather than CRLF is probably not going to end well. | 14:48.11 |
| | Not sure how VS copes with LF in source files. | 14:49.15 |
| | Rerunning with pgmraw.72 would rather defeat the purpose of a release test. | 14:50.04 |
| tor8 | Robin_Watts_: if it's only the vcproj files, we could add "*.vcproj text eol=crlf" to the gitattributes | 14:51.08 |
| | but it sounds like something else even fishier is going on from what you're | 14:51.21 |
| Robin_Watts_ | tor8: That might be reasonable, yes. | 14:51.24 |
| | but it leaves the submodules... | 14:51.34 |
| tor8 | saying about the submodules not working. | 14:51.34 |
| | which I have no idea how this could be affecting at all | 14:51.44 |
| Robin_Watts_ | Leave it with me to play for a bit. | 14:51.59 |
| | Thanks. | 14:52.01 |
| sebras | Robin_Watts_: I'm bisecting the bmp-thing. | 14:53.09 |
| tor8 | Robin_Watts_: set 'core.eol=crlf'? | 14:54.06 |
| Robin_Watts_ | for every submodule ? | 14:54.36 |
| tor8 | in the mupdf/.git/config should be enough to be inherited | 14:55.04 |
| | to the submodules | 14:55.07 |
| Robin_Watts_ | I tried that. Didn't seem to work. | 14:55.24 |
| tor8 | you may need to remove and reset all the files for changes to take effect | 14:55.44 |
| | rm -rf everything-except-".git" && git reset --hard | 14:56.06 |
| | && git submodule update | 14:56.20 |
| sebras | bmp changing colors slightly happened in 229761e8b7e8685620cf4c038d848bccd563aa12 the output intent callback was removed, perhaps that's it..? | 15:00.14 |
| Robin_Watts_ | sebras: I have no idea. | 15:02.18 |
| tor8 | sebras: sounds plausible. | 15:04.10 |
| Robin_Watts_ | What caused the diffs with the figures? Digits appear to have changed shape slightly. | 15:05.21 |
| tor8 | Robin_Watts_: font update. | 15:05.31 |
| | new URW fonts | 15:05.35 |
| Robin_Watts_ | Urgh. | 15:05.41 |
| | I'll run forwards through the bmpcmps if someone could run backwards and we'll meet in the middle. | 15:06.25 |
| | https://ghostscript.com/~regression/release__mupdf/compare11.html#519 - annotation difference? | 15:07.58 |
| tor8 | I'll go backwards then | 15:08.04 |
| | Robin_Watts_: yes. all the big speech bubble diffs are annotation differences. | 15:08.25 |
| Robin_Watts_ | shouldn't have just vanished though, right? | 15:08.38 |
| tor8 | it went properly-sized | 15:08.45 |
| | and changed icon to the three lines (since we now support more than just the speech bubble icon) | 15:08.59 |
| Robin_Watts_ | tor8: Cool. | 15:09.10 |
| sebras | tor8: fts_19_1903.pdf.ppmraw.300.0 on page 40 is presumably a new stamp annotation? | 15:12.02 |
| tor8 | yes. stamp annotations. | 15:12.48 |
| Robin_Watts_ | Everything up to and including page 15 looks OK. | 15:14.09 |
| | Everything up to and including page 25 looks OK. | 15:14.11 |
| tor8 | Down to 63 looks OK (I'm assuming broken image streams changing color from all white to all black is irrelevant) | 15:15.01 |
| sebras | I did check the bmp color against what it looked like back when I added the support. apparently something broke it and tor mostly unbroke it. | 15:15.03 |
| tor8 | sebras: 1555: tests_private/pdf/sumatra/use-after-free_from_pdf_load_compressed_stream.pdf.pgmraw.72.0 on page 65 | 15:15.35 |
| sebras | tor8: I think that was a j2k, right? I deemed that to be ok. | 15:15.37 |
| Robin_Watts_ | tor8: Do we render more or less like acrobat than before ? | 15:15.41 |
| tor8 | acrobat throws an error "insufficient data for image" and bails | 15:16.28 |
| | evince draws it black | 15:16.37 |
| | gs draws it white | 15:16.40 |
| | I think it's down to what we initialize the image color with before we bail decoding | 15:17.05 |
| sebras | that's a jbig2 that fails | 15:17.09 |
| | I think the default background color depends on some segment in it. | 15:17.28 |
| Robin_Watts_ | One of the hindi(?) fonts in ManyLang.epub (test 745) has changed from sans to serif | 15:18.13 |
| | Similarly for hebrew. | 15:18.51 |
| tor8 | Robin_Watts_: yes. same for the CJK fonts. | 15:19.27 |
| Robin_Watts_ | And all JP2 changes are progressions, as far as I'm concerned. | 15:20.28 |
| | page 35 is good. | 15:21.26 |
| sebras | tor8: https://ghostscript.com/~regression/release__mupdf/compare43.html fts_32_3228.pdf.pgmraw.300.1 is the location change of the strikethrough correct? | 15:23.05 |
| tor8 | Down to 55 is good. | 15:23.43 |
| moolc | tor8: kinda funny to see a release(at least in CHANGES) of mupdf having the same 1 4 as freshly released macOS, touting all that nice features only available in gl version.. the same gl that cider folk just deprecated | 15:23.46 |
| tor8 | sebras: yes. it was upside-down before. | 15:24.24 |
| sebras | fts_33_3309.pdf.pgmraw.72.0 seems to have removed two boxes and added a new one (presumably annotations). unknown why. | 15:24.30 |
| tor8 | sebras: the three boxes in 3309 are widgets | 15:25.19 |
| | two buttons (that we don't synthesize appearances for) and a text field (that we do) | 15:25.47 |
| sebras | yes, I just ran mutool. | 15:26.14 |
| tor8 | reverse situation from before, when we didn't synthesize the border on the text field and made a half-assed attempt at synthesizing button appearances | 15:26.19 |
| sebras | then that's acceptable I believe. | 15:26.35 |
| tor8 | it's expected at least :) | 15:26.42 |
| | synthesizing button appearances is on the TODO (but low priority, since most if not all proper files have button appearances already) | 15:27.04 |
| sebras | whey! the diff off /fts_34_3404.pdf.ppmraw.300.0 on page 43 is.. impressive. | 15:27.08 |
| | I get dizzy. :) | 15:27.14 |
| tor8 | Down to 50 is good. | 15:29.31 |
| Robin_Watts_ | Up to 40 is good. | 15:30.26 |
| sebras | there are lots of annotation differeneces in 42 and 43. | 15:30.48 |
| tor8 | page 48 case 1262 is missing a button appearance (which was just showing the thorn+ydiaeresis UTF16 BOM as latin-1 text) | 15:31.06 |
| Robin_Watts_ | sebras: Yes, but all look plausible. | 15:32.20 |
| | up to page 45 looks good. | 15:32.24 |
| tor8 | and down to 46 looks good | 15:32.53 |
| Robin_Watts_ | We're done then. | 15:33.03 |
| | And there were no diffs after the freetype release. | 15:33.26 |
| sebras | Robin_Watts_: there should be three files that differed. | 15:33.46 |
| Robin_Watts_ | I have no diffs showing with -w3 -t32 | 15:34.09 |
| sebras | Robin_Watts_: there arguments mean almost nothing to me. :) | 15:34.32 |
| | what is -w3? | 15:34.35 |
| tor8 | the website now has links to the RC tarballs | 15:34.37 |
| | https://mupdf.com/release_history.html and https://mupdf.com/downloads/index.html | 15:35.10 |
| | moolc: yes. screw the fruit company... | 15:35.25 |
| Robin_Watts_ | -w3 = allow for slight shifts in position of things (within a 3x3 window) | 15:35.55 |
| tor8 | dinners | 15:36.19 |
| moolc | tor8: screw them pedal to metal style | 15:36.27 |
| moolc | ditto | 15:36.31 |
| sebras | Robin_Watts_: ah. | 15:37.47 |
| moolc | https://sympa.inria.fr/sympa/arc/caml-list/2018-09/msg00044.html (better than sebras (or at least almost as good) at reviewing certain misfeatures) | 15:38.53 |
| Tamir_Evan | When I try to download the Windows binary archive for 1.14.0-rc1 (mupdf.com/downloads/archive/mupdf-1.14.0-rc1-windows.zip , linked from mupdf.com/downloads/index.html ), I get a "404 Not Found" page. | 18:03.55 |
| sebras | tor8: there's a number of things I found today on sebras/tried. | 19:35.12 |
| | tor8: "throw-upon-out-of-range-page" seems to be due to a misalignment between different document type about how to handle out of bounds page numbers. most throw and some code assumes this. | 19:36.54 |
| tor8 | Tamir_Evan: try now. | 21:00.08 |
| | Forward 1 day (to 2018/09/26)>>> | |
Log of #mupdf at irc.freenode.net.