| <<<Back 1 day (to 2018/05/02) | 20180503 |
dp | can I load pdf from url using mupdf? | 04:24.47 |
Robin_Watts | tor8: Hey. | 09:40.35 |
tor8 | Robin_Watts: morning. | 09:40.47 |
Robin_Watts | So, is MuPDF supposed to build on windows at the moment? | 09:40.48 |
| When I try to build mupdf-core, I get told "Can't find ..\..\source\cb\mutiff.c' | 09:41.36 |
| When I try to build mupdf-core, I get told "Can't find ..\..\source\cbz\mutiff.c" | 09:41.46 |
| and when I try and build mupdf-fonts, it can't find any of the dumped font files. | 09:42.05 |
kens | Oh, that's a regression in the last few days, mine worked a little while back but I haven't pulled the latest commits yet | 09:42.39 |
tor8 | commit 229761e8b7e8685620cf4c038d848bccd563aa12 "Handle multi-page image formats in muimg document driver." removes the mutiff.c file from the vcproj | 09:42.50 |
| and the make generate step is supposed to invoke the scripts\fontdump.nmake file | 09:43.20 |
Robin_Watts | tor8: Where does our vcproj live now? | 09:44.11 |
| I thought it was build/vs2005/ | 09:44.17 |
tor8 | same place as always, the build/vs2005 stuff is from the premake experiments | 09:44.28 |
Robin_Watts | platform/win32/ is the old (premake one) | 09:44.31 |
| Ah. | 09:44.33 |
tor8 | we haven't moved forward on premake yet ... still waiting to see how responsive they are to your patches | 09:44.51 |
Robin_Watts | Hmm. /nologo is giving my nmake conniptions | 09:48.11 |
tor8 | Robin_Watts: oops. | 09:49.11 |
| just remove the \nologo (I think my turn / into \ was a bit too aggressive) | 09:49.29 |
Robin_Watts | yeah /nologo works better. | 09:49.49 |
tor8 | Robin_Watts: top commit on tor/master should fix that | 09:50.22 |
| I guess I didn't catch that because I'd already successfully built hexdump.exe when I tried :( | 09:50.49 |
Robin_Watts | /nologo would work rather than \nologo | 09:51.09 |
| but lgtm. | 09:51.15 |
| oh, wait... | 09:51.23 |
tor8 | the /nologo would be messed up by the sed script that follows | 09:51.36 |
| or rather the 'tr' | 09:52.00 |
Robin_Watts | ok. | 09:52.08 |
| fatal error C1083: Cannot open source file: '..\..\generated\resources\fonts\noto\NotoSansGujarati-Regular.ttf.c': No such file or directory | 09:52.18 |
moolc | tor8: what's the point of echo -e when you have /bin/printf? | 09:52.27 |
| [-e not being in POSIX notwithstanding] | 09:53.24 |
Robin_Watts | tor8: generated\resources\fonts\noto\NotoSerifThai-Regular.otf.c != .\..\generated\resources\fonts\noto\NotoSerifThai-Regular.ttf.c | 09:54.18 |
| otf != ttf | 09:54.21 |
tor8 | Robin_Watts: hmm. I may have forgotten to update the libresources.vcproj | 09:54.50 |
| when I sucked in the new fonts | 09:54.58 |
| Yeah. oops. | 09:55.16 |
| I should regenerate the file list for libresources.vcproj from scratch | 09:55.37 |
Robin_Watts | ok. | 09:55.41 |
| I shall return to swearing at gs. | 09:55.50 |
| tor8: basically I was hoping to build and to come up with some sort of answer to give Raed. | 09:56.07 |
tor8 | on tor/win32 | 09:57.30 |
| okay, I pushed the fix to origin. win32 build should be back to working now. | 10:00.41 |
Robin_Watts | tor8: Ok, so libpkcs7 | 10:04.18 |
| In build events... | 10:04.27 |
| Prebuilt event is set to call generate.bat and "Generate CMap and Font source files" | 10:04.59 |
| For all the release ones. | 10:05.57 |
| and in lubmupdf | 10:06.25 |
tor8 | oh, now I found it. only in the 'release' mode. | 10:08.04 |
| shouldn't be necessary | 10:08.36 |
Robin_Watts | Yeah, I've stripped them out here. | 10:08.47 |
tor8 | it has 'generate' as a project dependency | 10:08.49 |
Robin_Watts | There are many instances. | 10:08.54 |
| It should have generate as a project dependency, I think. | 10:09.04 |
tor8 | libmupdf.vcproj has generate.vcproj as a build project dependency | 10:09.24 |
kens | Crumbs, loads of seg faults tor8 | 10:11.56 |
| Well, asserts or seg faults | 10:12.13 |
tor8 | kens: which file? | 10:13.21 |
kens | No the last regression run | 10:13.29 |
tor8 | oh... rats. | 10:13.36 |
kens | fixc /nolog | 10:13.46 |
tor8 | kens: I think they started with "Handle multi-page image formats in muimg document driver." | 10:15.16 |
kens | Em possibly | 10:15.25 |
| I may not have noticed that one, just saw this one go past | 10:15.35 |
tor8 | at least it's restricted to crashing on plain image files only and not PDF files... | 10:16.01 |
kens | Well that would seem likely to be the problem then | 10:16.23 |
Robin_Watts | tor8: http://git.ghostscript.com/?p=user/robin/mupdf.git;a=commitdiff;h=aed7cd7e87a387aa5ef4603436b30d60d0547bd1 | 10:16.51 |
tor8 | Robin_Watts: LGTM. | 10:17.19 |
Robin_Watts | kens: Right, so the contents for page 1 of Raeds file is a flate compressed stream, and the Length is given as 2. | 10:24.01 |
kens | Yes I think that's what I saw | 10:24.12 |
| It *is* wrong | 10:24.17 |
Robin_Watts | I think the answer is "don't use such broken files". | 10:24.17 |
kens | Except that both Ghostscript and Acrobat will open it | 10:24.31 |
Robin_Watts | gives a clarksonesque shrug. | 10:24.43 |
kens | Well, that's what we have to deal with all the time | 10:26.06 |
| If Acrobat will open it, then its not a problem file. | 10:26.15 |
| He could, of course, preprocess such files by running them through Ghostscript and the pdfwrite device, then it'd be OK | 10:26.35 |
Robin_Watts | We honour the length field. I'm not sure that it's right not too. | 10:27.35 |
kens | Well celarly Acrobat doesn't. Ghostscript checks it and warns that its not correct | 10:27.56 |
| We have *lots* of test fiels with incorrect Stream lengths I think | 10:28.13 |
| Obviously its up to you and tor | 10:28.35 |
Robin_Watts | I don't see it as a high priority to fix it. tor8 may feel differently. | 10:28.57 |
kens | As long as one of you replies to Raed, not my problem :-) | 10:29.14 |
Robin_Watts | I guess we should open it as a low priority bug for future consideration. | 10:29.20 |
tor8 | it would be non-trivial to fix well | 10:29.58 |
| we'd have to have some way of syncing all the streams that are chained and check if they reach EOF at the same time | 10:30.22 |
Robin_Watts | tor8: Where we have "fz_open_null(ctx, file_stm, len, offset)" | 10:30.31 |
tor8 | and allow the null filter at the penultimate location keep reading past its bounds | 10:30.43 |
Robin_Watts | we could make that fz_open_stream_terminated_by_stream. | 10:30.47 |
| and have that stream (when it reaches the specified len), check for the following bytes being "stream". | 10:31.12 |
tor8 | one way could be to let the null filter read until it hits its len, and then if you ask for more data, it would start scanning for the 'endstream' keyword | 10:31.23 |
Robin_Watts | If not, print a warning and keep going. | 10:31.24 |
tor8 | and print a warning if you go past the original end | 10:31.33 |
Robin_Watts | yeah, I meant "endstream", not "stream" | 10:31.42 |
| I think we're in agreement. | 10:31.47 |
tor8 | Robin_Watts: yeah, I think we're saying the same thing :) | 10:31.51 |
Robin_Watts | I'll open a bug and reply to Raed, unless you want to. | 10:32.08 |
tor8 | I'm happy to leave this to you | 10:32.17 |
| Robin_Watts: okay, so fix for the "oops I broke the cluster" on tor/master | 10:33.23 |
| and another few commits that could do with a lookover | 10:33.29 |
Robin_Watts | That one lgtm. | 10:33.55 |
tor8 | all the old cross compiler targets/configurations ... I removed the ones that I think are no longer used | 10:34.20 |
| are you still using the tizen and beagleboard ones? | 10:34.32 |
Robin_Watts | and the others look fine to me. | 10:37.04 |
| I am not. | 10:37.06 |
| But having them there doesn't hurt as examples. | 10:37.25 |
tor8 | Robin_Watts: I think having the mingw32 sections would be enough as examples | 10:38.13 |
Robin_Watts | ok. | 10:45.43 |
moolc | someone uses cscope while developing on mupdf? (Makefiles seems to suggest that, but maybe it's just some stale thing?) | 12:47.45 |
tor8 | moolc: I used to, but I've settled for regular ctags | 12:50.38 |
moolc | tor8: a coworker (vim user btw) showed me that git grep is actually somewhat superior alternative for all practical purposes | 12:51.30 |
tor8 | moolc: git grep doesn't allow for ^] | 12:51.57 |
| and yes, I do already have "set grepprg=git\ grep\ -n" in my .vimrc | 12:52.12 |
moolc | tor8: ^] to abort or something? | 12:52.28 |
tor8 | ^] goes to the definition of the symbol you're standing on | 12:52.40 |
moolc | ah.. shows my vimgnorance | 12:52.56 |
| "C-] runs the command abort-recursive-edit (found in global-map), which is an interactive built-in function in âC source codeâ." in emacs, hence my confusion | 12:54.43 |
| | 12:54.43 |
avih | fwiw, if you don't mind gui, i find vscode quite good an analyzing and navigating projects by symbols definition, declaration, etc. | 12:56.50 |
| s/an/at/ | 12:57.02 |
| Forward 1 day (to 2018/05/04)>>> | |