| | <<<Back 1 day (to 2013/12/31) | 2014/01/01 |
| Robin_Watts | sebras: ping ? | 11:10.20 |
| | sebras: I think you had something to do with the SHA routines in MuPDF, right? | 13:26.01 |
| | I've got a commit online that tweaks them a bit. | 13:26.18 |
| | The primary intent is to avoid unaligned accesses, but I've also tweaked them a bit to avoid some copying. | 13:27.14 |
| sebras | Robin_Watts: mmm, I think I took zenikos patches and added them to mupdf. | 13:47.10 |
| | Robin_Watts: I've had a look at the patch now and it LGTM. | 13:57.01 |
| | Robin_Watts: at first I thought that you might need to update blk0() as well, but as you change the argument name then that is of course not necessary. | 13:57.39 |
| Robin_Watts | sebras: Yeah, I don't change any of the central garbling calls, just the setup for it. | 14:03.45 |
| | Thanks. | 14:03.53 |
| | sebras: So, todays random thought for MuPDF. | 14:31.22 |
| | zeniko has an open bug complaining that memory use goes up a lot if you search through a file, because we end up caching lots of objects in memory. | 14:31.51 |
| | tor suggested a 'pdf_clear_xref' function that would run through the xref and bin all the object references (presumably only those where refs == 1, so only the xref table is holding the refs). | 14:32.47 |
| | zeniko disliked that because it might bin too much. | 14:33.22 |
| | I think he also disliked it because it loads and caches everything only to discard it again. | 14:33.52 |
| henrys | Robin_Watts: I was suggesting you and Paul might spend some small amount of time with the code, the priority is still mupdf and presumably the full timer would do most support. | 14:36.12 |
| Robin_Watts | henrys: right. | 14:36.28 |
| | henrys: I'd be really wary about trying to make money from app sales. | 14:38.32 |
| | I mean, doing a build and putting it out there seems sensible given what we have, but it's hugely difficult to do any sort of big development and hope to get realistic money back. | 14:39.09 |
| henrys | Robin_Watts: I would strongly agree if we didn't have a completely finished product dumped in our lap sitting on a server. | 14:39.34 |
| Robin_Watts | This is 4 years old now, and the situation has only gotten worse: http://www.streamingcolour.com/blog/2009/03/09/the-numbers-post-aka-brutal-honesty/ | 14:40.38 |
| henrys | Robin_Watts: a lot rides on the quality of the code. | 14:41.26 |
| | Robin_Watts: freemium (printing & editing) is a thought I had last night. | 14:42.58 |
| Robin_Watts | I chatted briefly to Alex this morning. | 14:43.14 |
| | Smart Office do a free 'lite' version, and a paid for 'pro' version. | 14:43.32 |
| | free version can't save or print. | 14:44.22 |
| | and limits the number of pages you can view. | 14:44.57 |
| | hmm. maybe it can print | 14:46.05 |
| sebras | Robin_Watts: couldn't we somehow tag all objects that we have in the xref before searching and then clear anything that is unmarked after searching? | 14:46.13 |
| | Robin_Watts: that would alleviate zenikos claim | 14:46.22 |
| Robin_Watts | sebras: That was what I was building to :) | 14:46.31 |
| sebras | Robin_Watts: but it would take some time for a BIG pdf like pdfref. | 14:46.33 |
| Robin_Watts | I'd worry that we might end up reloading font details for every page. | 14:47.05 |
| sebras | Robin_Watts: we have the flags in pdf_obj, so that could be used I guess. | 14:47.18 |
| Robin_Watts | i.e. suppose we have a font that's only used on pages 10-1000 | 14:47.45 |
| | We load the doc, and show page 1, and process page 2 to get the thumbnail. | 14:48.01 |
| sebras | Robin_Watts: I guess zeniko is also thinking about seeking for a text in a big document (maybe with many fonts shared between pages) where you can't find the text anywhere. | 14:48.17 |
| | and then we leave all the resources cached unnecessarily. | 14:48.26 |
| Robin_Watts | Then when we search if we run each page to do the search and then discard each page as we go, the font will be loaded (for it's tounicode map) 9990 times. | 14:48.59 |
| sebras | I haven't really read his bug report, but I think I understand the scenario. | 14:49.05 |
| | Robin_Watts: you don't need to clear the unused objects for every page. | 14:49.34 |
| | you could decide to do it for every 10 pages or some arbitrary number | 14:49.46 |
| henrys | Robin_Watts: more recent: http://www.forbes.com/sites/tristanlouis/2013/08/10/how-much-do-average-apps-make/ the average's are poor but this could be a significant product. | 14:49.54 |
| Robin_Watts | That reminds me of the lie that all parents tell one another to be able to procreate. "Our Child Will Be Different" :) | 14:52.14 |
| Robin_Watts | pops afk for a few hours. Back later. | 15:03.23 |
| henrys | Wow surreal - marijuana for sale today - licensed retail stores. Unbelievable. | 15:06.30 |
| Robin_Watts | denvers productivity drops off.... | 15:09.01 |
| | cheeto consumption increases. | 15:09.14 |
| | Forward 1 day (to 2014/01/02)>>> | |
IRC Logs