| <<<Back 1 day (to 2020/07/05) | Fwd 1 day (to 2020/07/07)>>> | 20200706 |
ator | paulgardiner: yeah, you don't want -z to get a readable file (it recompresses the streams to be unreadable again). mutool clean -dif -a -c are the options to get the most readable output (dif decompresses all but leaves images and font files compressed), -a asciihex encodes binary data (the fonts and images mostly) and -c pretty-prints content streams | 08:53.06 |
| avih: thanks! I'll get that patch committed. | 08:53.39 |
| avih: good job finding it! | 08:53.48 |
avih | ator: thx. fwiw, mpv does use iterators a lot (every object used at the mpv api is converted to an mpv data node with an iterator), but not every iterator gc causes it. it only happens if the same target is inserted to the list while it's there - which is not always the case. | 08:59.56 |
| i.e. even if it's already marked it won't necessarily cause an issue, but it might. | 09:00.53 |
malc_ | avih: btw have you tried Fabrices JS implementation? | 09:01.29 |
avih | malc_: define "tried" | 09:01.46 |
malc_ | avih: okay let's reword - have you looked into it | 09:02.27 |
avih | yes | 09:02.32 |
malc_ | verdict? | 09:02.38 |
avih | early impression: complex code, hard to use API. apparently relatively fast and complete. | 09:03.11 |
malc_ | gotcha thanks for the summary | 09:03.34 |
| it has one giant plus compared to mupdf though | 09:04.25 |
paulgardiner | ator: thanks. I've taken a note of that for another tiem | 09:05.29 |
avih | for me the main advantage over mujs is the completeness for ES6. speed is nice but depending on context, might not matter much (in mpv it doesn't matter). but the API is really hard to use. | 09:05.44 |
malc_ | for me lack of tabs is the main selling point | 09:06.19 |
avih | also, the code is not easy to hack. | 09:06.23 |
| tabs? | 09:06.29 |
ator | ESnew (6, 2015, 2017, Harmony, whatever they're calling it today) is another completely different language, sadly :/ | 09:06.37 |
malc_ | yeah... you know character with code 9 | 09:06.50 |
avih | ator: yeah, but people do use it. | 09:06.56 |
malc_ | ator: how so? | 09:07.16 |
avih | personally i'm fine with 5.1 for mpv, i think it's perfectly enough and well defined, but ESnew is more popular these days | 09:07.25 |
ator | it adds soooo much extra stuff, if I were to implement it I'd basically have to rewrite big chunks from scratch | 09:07.48 |
malc_ | ator: also got my e-mail? | 09:07.50 |
| things that have humble beginnings usually do grow into monsters... oh well | 09:08.29 |
| we all know now the fate of mupdf | 09:08.40 |
ator | malc_: haven't seen an email from you in a while. what's the subject line? | 09:09.03 |
malc_ | ator: sec | 09:09.16 |
avih | duh.. you need to remove him from your spam filter... | 09:09.23 |
ator | malc_: oh, the monitor question? | 09:09.44 |
malc_ | ator: yeah that's the one | 09:10.38 |
ator | windows and cocoa render at twice the resolution, but multiply all coordinates by 2 before rendering IIRC | 09:11.05 |
| well, apple is a bit worse than that. it renders at 2x the normal resolution and *then* scales to fit the actual monitor | 09:11.22 |
malc_ | avih: i have a gmail address, i'm impervious to spam filtering! | 09:11.23 |
ator | which on some devices is like 1.75 or something | 09:11.36 |
| so you get blur no matter what on an apple retina device | 09:11.43 |
malc_ | insane | 09:11.47 |
avih | malc_: it's not up to you... | 09:11.47 |
malc_ | avih: how do you possibly know that i'm marked as spam in emails sent to ator? | 09:12.27 |
ator | malc_: *but* if a program on windows doesn't say it's capable of hidpi, then it's rendered at 1x and scaled up | 09:12.30 |
avih | ator: btw, if you don't take my patch as is, please do use a meaningful commit message which specifies at least: 1. the issue in practice. 2. which commit caused the regression in practice. | 09:13.03 |
malc_ | ator: i really really prefer the Xft.dpi approach even though it sucks too | 09:13.18 |
ator | avih: I had to fiddle a bit to get 'git am' to accept it, but I've taken it with your full analysis comment in place, don't worry :) | 09:13.42 |
| malc_: everything sucks: a concise history of software. | 09:14.23 |
avih | ator: really? i tried it from that post, copied it entirely, then xclip -o | git am and it worked cleanly | 09:14.33 |
malc_ | ator: you know what sucks even more? | 09:14.42 |
ator | avih: something messed up whitespace when copy-pasting it from gmail | 09:14.56 |
malc_ | ator: btw was my analysis of the scaling problem close to truth btw? | 09:15.02 |
ator | malc_: tabs? | 09:15.09 |
avih | gotcha. i copied from github | 09:15.13 |
malc_ | ator: distant second, dyson | 09:15.22 |
ator | malc_: yeah. old software gets the heavy handed "render at low-res then scale up" treatment | 09:15.46 |
| malc_: this is mostly theoretical knowledge, I don't own a high dpi device (yet...) | 09:16.27 |
malc_ | ator: tusen tack! | 09:16.31 |
ator | malc_: IMO everything should have taken the Xft.dpi or equivalent thing and scaled the UI including other graphical elements by the same amount | 09:17.22 |
| and based the size of menus and scrollbars by the font size | 09:17.33 |
| but that ship sailed back in the early 90's or so... | 09:17.41 |
malc_ | ator: yeah :( | 09:18.02 |
ator | Java tried to do the right thing with adjustable layout managers in their AWT and Swing toolkits | 09:18.22 |
| but win32 and macos classic and macos x all based their GUI layouts on absolutely positioned and sized elements. | 09:18.46 |
malc_ | shrug | 09:20.35 |
avih | ator: also, i was able to reproduce random errors relatively consistently (as in with some usage pattern it would typically happen after a minute or so at most), and since i applied that patch i could not reproduce it anymore, and it never happened again. | 09:21.02 |
malc_ | https://tonyxzt.blogspot.com/2010/01/tests-can-show-presence-of-bugs-not.html | 09:23.23 |
| ator: not sure if i ever thrown https://www.youtube.com/watch?v=8IzVYIkzw8E at you, so here goes.. | 09:26.26 |
ator | malc_: eek! like early 90's before people got wind of things like encodings and unicode :) | 09:35.26 |
| or should I say late 80's... | 09:35.32 |
malc_ | ator: not a fun of Symbols font i take it :) | 09:35.56 |
| fan | 09:36.17 |
ator | a prime example! and wingdings/dingbats too. | 09:36.26 |
malc_ | damn you homo-phony damn you to hell | 09:36.37 |
| ator: both symbols and beograd are EXCELLENT tools to learn Greek and Cyrillic alphabets though :) | 09:37.53 |
| or rather boe.. cry and symb... grek are e... | 09:38.18 |
| cry=cryl | 09:38.28 |
| damn my inability to type continues | 09:38.46 |
| and i practice keyhero and 10fastfingers daily | 09:39.00 |
ator | malc_: yeah. maybe I should do that one day... | 10:10.28 |
malc_ | ator: you want to suck just like me (typing wise) okay.. jedem das zeine | 10:12.48 |
ator | malc_: meant using a fake-characters-font to get more used to reading greek/cyrillic alphabets | 10:13.19 |
| I'm happy enough with my typing :) | 10:13.34 |
malc_ | ator: i know you speedy gonzales you... | 10:13.46 |
| and yeah my greek comprehension really improved | 10:13.55 |
| i replaced linux libertine with symbol (https://github.com/moosotc/snippets/blob/master/.config/fontconfig/gen.sh) and now all the headings in wikipedia articles are in pseudo greek | 10:14.57 |
avih | ator: btw, i think this is worthwhile - use proportional GC threshold instead of fixed. first commit: https://0x0.st/iyV3.txt second: https://0x0.st/iyVY.txt | 11:07.33 |
| it makes the impact of GC much more consistent, and the "predictive" aspect it works extremely well, even with changing use pattern over time | 11:08.19 |
malc_ | avih: evidently your .st freundin was unable to fix the Tor issue | 11:09.57 |
| shame | 11:10.01 |
avih | malc_: what are you talking about? | 11:10.17 |
| oh, 0x0? | 11:10.29 |
malc_ | 0x0.st does not work with Tor (the onion router) | 11:10.44 |
avih | i'm sure you can work around it. | 11:10.46 |
| but when i ask you to look at pastes, i'll use a different one. | 11:11.12 |
malc_ | appreciated | 11:11.27 |
| and yes things work when i use (say) links, but that's just jumping through hops | 11:12.03 |
avih | good to keep in shape! | 11:19.11 |
| ator: i didn't try to apply this patch after the recent gc de-recurse, but it should be fairly trivial to do. let me know if it doesn't apply well. | 11:24.27 |
| these patches* | 11:24.38 |
| for me the first patch applied cleanly, and the second patch applied correctly with git-auto-fix (i think `git am -3' ?) | 11:27.54 |
| that's the second patch applied to master+1st-patch: https://0x0.st/iyV7.txt | 11:29.41 |
malc_ | avih: with tor browser 0x0 is accessible (even though after a *very long* pause), with standalone tor (used as a scocks5 proxy things) do not work at all | 11:33.29 |
avih | ator: the "Silence gcc warning about overflow." commit is rather superficial, is it not? wouldn't it be better to use unsigned char instead? | 11:33.41 |
malc_ | ator: part of the reason for the stuff showcased in the 'perils of the fonts past' youtube video is showcased here: https://boblycat.org/~malc/scratch/wfs.png designers use insane font stacks... | 12:53.05 |
ator | malc_: insane, or copy pasted from frameworks they haven't a clue what's going on? | 12:54.32 |
malc_ | ator: current web is full of this... | 12:54.53 |
| with begrad i can at least see it.. | 12:55.07 |
| and look elsewhere | 12:55.11 |
| ator: i think a) parts of the world are to windows centric b) leave in the past c) clueless d) switching to macos (and that scares the living shit out of me) | 12:56.13 |
avih | ator: any comment on the proportional gc patches/approach/idea? | 13:00.08 |
| <<<Back 1 day (to 2020/07/05) | Forward 1 day (to 2020/07/07)>>> | |