| <<<Back 1 day (to 2018/11/13) | 20181114 |
sebras | tor8: how about adding "Bug 699523: Remove declaration of removed function."Bug 699523: Remove declaration of removed function." from sebtras/master? | 13:04.53 |
Robin_Watts | sebras: Just looking at warnings in gs again. | 13:09.35 |
| c:\artifex\ghostpdl.git\jbig2dec\jbig2_halftone.c(79) : warning C4146: unary minus operator applied to unsigned type, result still unsigned | 13:09.37 |
| Do you know about that one? (Dio you try to patch that before and hit problems?) | 13:09.53 |
sebras | not sure. | 13:10.01 |
Robin_Watts | tor8: The jpegxr stuff in gs... are we free to tweak that to fix warnings, or is that asking for problems? | 13:10.17 |
| In particular I'd like to change DEBUG(...) to be JXR_DEBUG(...) | 13:11.51 |
| but if you're going to be importing new versions, that'll probably conflict horribly. | 13:12.11 |
sebras | Robin_Watts: I think that's the reference implementation, unlikely to change. | 13:12.59 |
tor8 | Robin_Watts: I haven't looked if there's a new one in agesn | 13:14.27 |
| the code is a right mess | 13:14.39 |
Robin_Watts | tor8: Any objection to me making that change? | 13:16.41 |
| We can always do the same rename in any new version we want to import. (or rename it back) | 13:17.19 |
tor8 | Robin_Watts: https://jpeg.org/jpegxr/software.html might be worth trying to upgrade first? | 13:17.53 |
| 1.41 is the newest tarball | 13:18.43 |
Robin_Watts | Do we actually use this stuff? | 13:18.45 |
tor8 | mupdf's thirdparty is at 1.32 | 13:18.52 |
| looks like ghostpdl is still using 1.09 | 13:19.00 |
Robin_Watts | We use this in mupdf?! | 13:19.06 |
tor8 | for XPS | 13:19.13 |
| but I think we've turned it off by default, and only build if it's present | 13:19.33 |
Robin_Watts | It ain't in my mupdf thirdparty. | 13:19.36 |
tor8 | if you've checked it out, we can build with it | 13:19.58 |
| but seeing how shoddy the code is, I don't want to enable it by default | 13:20.05 |
Robin_Watts | right, so all the windows releases we've done haven't had it in. | 13:20.11 |
tor8 | god knows how many exploits lurk inside | 13:20.14 |
Robin_Watts | tor8: That git whitespace fix thing you gave the other day doesn't fix any commits for me. | 13:21.34 |
| All it does is screw makefiles whenever I rebase :( | 13:21.46 |
tor8 | you need the proper .gitattributes for it to work, I think | 13:22.26 |
Robin_Watts | If you have some git magic that tells git to ignore that rule for makefiles then it'd be great to get that into gs. | 13:22.29 |
tor8 | take a peek at mupdf's .gitattributes | 13:22.40 |
| that has filetype specific whitespace rules | 13:22.52 |
sebras | tor8: they've even updated on the itu-t website. I wasn't aware. | 13:22.56 |
Robin_Watts | tor8: I see. | 13:24.05 |
tor8 | wow! "Note that due to this change | 13:24.20 |
| the API of the reference software also changed slightly. On decoding, | 13:24.21 |
| the "pixel format" is no longer available, instead descriptive | 13:24.21 |
| functions have been added that identify the format by the number of | 13:24.21 |
| channels, bits per channel, color space etc." | 13:24.21 |
| they've actually made the API usable?! | 13:24.26 |
moolc | tor8: it seems you have introduced some set but not used warnings lately | 13:35.39 |
tor8 | moolc: not intentionally | 13:39.32 |
moolc | tor8: if i made that implication it wasn't intentional either | 13:43.41 |
sebras | moolc: where are the warnings your talking about? | 13:44.10 |
tor8 | svg- | 13:44.16 |
| run.c has one | 13:44.18 |
| the 'rtl' one in stext-device is there to remind me to fix some RTL ordering stuff | 13:44.36 |
moolc | sebras: i can rerun the build (it's ccached anyway) or you can just try "gcc version 8.2.1 20180831 (GCC)" and native build | 13:45.45 |
moolc | people do use a decent 3 Ð friendly typeface it seems https://arstechnica.com/gadgets/2018/11/the-next-version-of-http-wont-be-using-tcp/ | 13:47.35 |
Robin_Watts | tor8: I should probably not get sidetracked into that new jpegxr version thing, especially as it doesn't look like it'll be a drop in replacement. | 13:48.23 |
tor8 | I'll put it on my TODO list | 13:49.15 |
moolc | sebras: https://boblycat.org/~malc/scratch/mupdf.build | 13:51.31 |
sebras | moolc: I see them. at least some are present in build=releas too. | 13:51.48 |
tor8 | sebras: commit to fix one of the set but unused functions on tor/master | 13:52.38 |
moolc | sebras: let me stress once again that realforce sucks.. decent keyboard would have stopped you from typing "build=releas" | 13:52.55 |
sebras | moolc: I mistype on every keyboard though... | 13:53.47 |
moolc | sebras: but decent keyboard would have flashed it backlights and applied some current to your naughty fingers... and eventually pavlovian forces would have made you to a) not bottom out b) stop misspelling | 13:55.22 |
sebras | tor8: the first two commits on tor/master LGTM./ | 13:55.32 |
| moolc: I get my pavlovian forces from tor8 when he scolds me for typing too loudly. :) | 13:56.23 |
moolc | sebras: :) btw i would have understood the misspellings if you typed danish but english! c'mon.. as we all know - https://www.youtube.com/watch?v=H1-GDlzODzc | 13:57.33 |
pink_mist | moolc: also see https://www.youtube.com/watch?v=s-mOy8VUEBk :P | 13:58.46 |
moolc | pink_mist: bashing danish is apparently NRK's "thing" https://www.youtube.com/watch?v=FqgRC5sfCaQ | 14:00.54 |
| but who can blame them.. 500 years of oppression | 14:01.07 |
pink_mist | lol | 14:03.05 |
sebras | tor8: ehm.. javah seems to have disappeared from openjdk since at least openjdk-10. | 14:12.18 |
| seems to have been done in january 2018 https://bugs.openjdk.java.net/browse/JDK-8182758 | 14:13.45 |
tor8 | sebras: oh dear. replaced with what? | 14:15.43 |
sebras | javac -h | 14:15.47 |
| so now the compiler itself outputs header-filers. | 14:15.55 |
| I'm trying to run it. | 14:15.58 |
| tor8: apparently we now get one header per class-file it seems. | 14:16.41 |
| tor8: on top of sebras/master you'll find the ugliest fix for supporting thig. | 14:19.35 |
| this. | 14:19.36 |
| tor8: what version of java do you run? | 14:22.19 |
tor8 | java version "1.8.0_121" | 14:22.42 |
| javac 1.8.0_121 | 14:22.46 |
sebras | tor8: I believe that it also supports the -h option. | 14:23.07 |
tor8 | sebras: it does, according to javac -help | 14:23.52 |
sebras | yup. | 14:23.56 |
| so, perhaps we ought to be using that instead. | 14:24.09 |
| tor8: I've forgotten. why do we check in mupdf_native.h? | 14:29.02 |
tor8 | sebras: for the android build. | 14:42.09 |
sebras | right. | 14:42.31 |
| tor8: well then, how do you like sebras/master? | 14:51.59 |
tor8 | replace "cat ../../build/java/*.h" with "cat $(ls ../../build/java/*.h | sort)" | 14:57.00 |
| you don't want the generated file to be dependent on the inode or other random file system ordering of wildcard globbing | 14:57.20 |
| or $(find ../../build/java -name '*.h' | sort) | 14:58.16 |
sebras | right. bash always sorts wildcards so I haven't noticed, but your right I probably shouldn't depend on that. | 14:58.41 |
moolc | tor8: ls -S doesn't cut it? | 15:09.48 |
sebras | moolc: ls -S sorts by file size, does it not? | 15:10.52 |
moolc | sebras: âS | 15:11.31 |
| Sort with the primary key being file size (in decreasing order) and the secondary | 15:11.32 |
| key being filename in the collating sequence (in increasing order). | 15:11.32 |
| point being that it doesn't depend on inode or anything like that | 15:12.18 |
tor8 | sebras: -z noexecstack is not mentioned in the commit message | 15:48.42 |
| moolc: but it would lead to unneccessary diff churn if one file suddenly grows larger than another | 15:48.58 |
sebras | tor8: ah, that should probably be a separate commit. | 15:49.09 |
| it fixes a warning that openjdk-10 emits at runtime. | 15:49.53 |
| clang appears to support the same flag. | 15:50.01 |
| there's also an external binary called execstack that can be used to post-process the .so-file. but I think you prefer this linker option. | 15:50.34 |
tor8 | sebras: hang on. try adding "LC_COLLATE=C sort" to that build step | 15:50.34 |
| I think your locale is messing up sorting | 15:50.48 |
moolc | tor8: true that | 15:50.55 |
sebras | tor8: what's the semantic difference between Quad and Rect? | 15:54.31 |
tor8 | sebras: a Rect is orthogonal and axis aligned | 15:55.00 |
| a Quad can be rotated and/or skewed | 15:55.08 |
sebras | tor8: ok, in that case I prefer to propagate Quads as far as possible and only convert to Rect when needed. | 15:57.11 |
| I of course encountered numerious build issues while attempting to build mupdf-mini to make sure that my mupdf_ | 15:59.26 |
| native.h works fine. | 15:59.29 |
| numerious?! numerous. stupid fingers. | 16:00.00 |
tor8 | sebras: yes, the android viewers need updating | 16:04.53 |
| preferably to not need to convert to Rects at all | 16:05.15 |
sebras | I know. I'm working on it. | 16:05.16 |
tor8 | by drawing the Quads as quads | 16:05.24 |
| cool. I'll leave you to it then :) | 16:05.36 |
sebras | I simply do Quad.toRect() for now. | 16:05.53 |
| I need to get this crap to compile first. :( | 16:06.00 |
moolc | Numerious Topras sebras' arch nemesis... | 16:28.57 |
| Forward 1 day (to 2018/11/15)>>> | |