| | <<<Back 1 day (to 2018/01/18) | 20180119 |
| malc_ | tor8: stumbled on a .pdf (made out of .svg by cairo) that's rather peculiar- it's rendering time is "exponentially" related to the resolution (takes 1.2,2.9,5.9,11.39 seconds for 100,200,300,400 -r arguments to mupdf-x11) https://boblycat.org/~malc/hm.pdf | 13:17.45 |
| kens | Probably has transparency | 13:18.14 |
| tor8 | malc_: your https certificate has expired | 13:19.18 |
| kens | LOL I was aboput to say that :-) | 13:19.29 |
| | Yep, the page has a Group | 13:20.16 |
| malc_ | tor8: yeah, i just notified the person qho qualifies "your" moniker in "your https certificate" better | 13:20.20 |
| tor8 | kens: I suspect it's due to clip masks as well, since we render all clips as soft masks for anti-aliasing | 13:20.33 |
| kens | And constant alpha CA and ca not 1 | 13:20.35 |
| | Defines a load of ExtGStates too | 13:21.11 |
| | 270 of them | 13:21.24 |
| | bunch of shadings | 13:21.34 |
| | hundreds of image XObjects | 13:21.47 |
| | Looks like each image XObject has its own Group too | 13:22.38 |
| | Each ExtGState has a SMask | 13:23.01 |
| | Basically its an amazingly complicated file | 13:23.21 |
| | Oh, looks like the Form XObjects mostly just point to other Form XObjects which just do a fille | 13:24.20 |
| malc_ | fwiw it was generated by rsvg-convert out of /usr/share/icons/gnome/scalable/mimetypes/libreoffice-oasis-spreadsheet-template.svg | 13:27.02 |
| kens | Makes GS think hard, debug build (which admittedly is very slow) takes 11 seconds at 200 dpi | 13:27.07 |
| | malc_ : just saying its a lot more complicated a file than you might think from looking at it. | 13:27.28 |
| | Part of the problem is Cairo | 13:27.35 |
| | When creating PDF it makes everything into a group, and everything gets assigned as transparent. Even if its actually opaque | 13:27.57 |
| | Presumably because that's how Cairo's imaging model works | 13:28.11 |
| malc_ | kens: no doubt.. apparently it kills firefox/pdf.js when zooming in | 13:28.14 |
| kens | But in PDF, transparency involves doing a *lot* of extra work | 13:28.23 |
| | Yeah I can't say I'm surprised :-) | 13:28.39 |
| | Doesn't seem too slow in my FF | 13:29.09 |
| | I'm up to 530% how far do I need to go ? | 13:29.41 |
| malc_ | kens: try zooming in | 13:29.42 |
| | oh | 13:29.44 |
| | kens: i don't know.. i have 3.7g of ram and no swap | 13:30.04 |
| | maybe it just runs out of memory | 13:30.10 |
| kens | I guess that's possible | 13:30.16 |
| malc_ | fwiw pdfim (nee foxit) manages it just fine, or so it seems | 13:30.35 |
| kens | One of my FF tabs is up to 1.4Gb | 13:30.38 |
| malc_ | i also have overcommit disabled and conkeror(firefox-esr) running alongside | 13:31.19 |
| | so | 13:31.20 |
| kens | Well that's 1000% I think I'll stop there :) | 13:31.24 |
| malc_ | kinda short on memory | 13:31.25 |
| | hah | 13:31.29 |
| kens | <sigh> today's tip; don't try to read a PDF file with an XPS interpreter.... | 13:35.26 |
| malc_ | https://en.wikipedia.org/wiki/Robustness_principle | 13:36.31 |
| Robin_Watts | malc_: Brittle code is best. | 16:30.41 |
| | That way we wouldn't be in the mess we're in with crap standards like PDF :) | 16:31.03 |
| fredross-perry | Peanut brittle is best. | 17:55.10 |
| Robin_Watts | peanuts are the work of satan. | 18:01.10 |
| titanous | sebras: let me know when you have a Google account to add to oss-fuzz, there are now a bunch of bugs found | 22:47.21 |
| | Forward 1 day (to 2018/01/20)>>> | |
Log of #mupdf at irc.freenode.net.