| | <<<Back 1 day (to 2020/06/26) | Fwd 1 day (to 2020/06/28)>>> | 20200627 |
| avih | ator: i still didn't get to add the detailed stack logging, but the issue keeps happening. i still can't reproduce, and it seems to happen in random scripts and random places at the code. i think the common to the scripts where it happens is that they all work "continuously" and poll things many times/sec, which i think could point to the gc commit, because these script would do gc more than others. | 14:58.53 |
| | (it happened about 4-5 more times since we talked) | 14:59.15 |
| | also, it only happens in js scripts in mpv, and i do also have some lua scripts which don't exhibit any issue, and also i haven't noticed issues outside of mpv, so so far i think it's a js issue. possibly at the mpv side, but there were not recent related changes on the mpv side, so i still think it's mujs | 15:03.44 |
| | so it would be nice if you could review the gc commit. i'll try too, but last time i looked at it, i didn't quite get how it worked. i did know how it worked before that de-recurse commit. | 15:21.02 |
| | also, the error "TypeError: object is not callable" seems to be recurring, though i've seen others too, but this specific one definitely happened few times, so i think this could also point to the gc commit | 15:30.55 |
| | and i think it's related to arrays. last time the error was on indexOf() of a a string array element. | 15:37.11 |
| | though it did also happen at least once unrelated to arrays, on `(0, t.callback)()' | 15:38.06 |
| | actually, it is related to arrays. t is t = somearray.pop() | 15:39.08 |
| | and all the elements have a callback member | 15:39.29 |
| | and the array was not empty | 15:40.02 |
| sebras | avih: ator will not be back next week. | 16:08.05 |
| avih | yeah, i know | 16:08.19 |
| | but he might read during the weekend, or just when he gets back | 16:08.36 |
| sebras | avih: sure. as long as you are patient it will work out. :) | 16:14.54 |
| avih | not like i have a choice :) | 16:15.23 |
| malc_ | avih: you can always pester sebras... | 16:37.53 |
| | just saying | 16:37.58 |
| sebras | malc_: yes, but you have to pester me _REALLY_ hard to make me delve into mujs. | 16:39.03 |
| malc_ | sebras: it's a miracle you don't have to push tor really hard to work on js, given his lua preference, and avih here tries to make js in mpv work just as good as lua... different strokes... i'spose | 16:40.36 |
| sebras | malc_: me wanting to learn how to do do an interpreter and us needing a js interpreter was actually the impetus for ator starting mujs. | 16:41.39 |
| malc_ | sebras: WOW.. you want something and tor does it... how quaint! | 16:43.09 |
| sebras | malc_: we started it together, but ator had more time to spend on it and I wasn't able to keep up. | 16:47.55 |
| | malc_: nowadays it's definitely his baby, not mine. | 16:48.14 |
| malc_ | sebras: gotcha... thing is the way i read it js is mandatory part of pdf spec (1.7) | 16:48.34 |
| | you Cuckoo еыа | 16:50.01 |
| sebras | malc_: doesn't mean mupdf used to implement it. | 16:57.56 |
| | malc_: transfer functions are probably also a mandatory thing, but mupdf doesn't have it. | 16:58.26 |
| malc_ | sebras: but you do try. no? | 16:58.31 |
| | for forms and whatnot | 16:58.35 |
| malc_ | has no idea what transfer functions are :( | 16:58.53 |
| sebras | malc_: for forms and annotations and digital signatures, yes. but there are other things that we haven't started on yet. | 16:59.08 |
| malc_ | luckily i do not care :) | 16:59.39 |
| sebras | malc_: pdfref17 6.3 | 17:06.10 |
| malc_ | i'm (praise be the gods) almost color blind | 17:08.10 |
| | so i can sorta feel your pain when dealing with cmyk lab littlecms and what not | 17:08.49 |
| | but on a strange level | 17:08.55 |
| avih | without further analysis, i think what happens is that some objects get freed incorrectly, but it keeps working because the memory content is still there. i think i'll try to make the gc free also zero or write some pattern to the memory before free, to make errors appear earlier. | 23:08.50 |
| | if it's a gc issue due to 331c5ecb, then it should expose it more clearly, i hope | 23:10.18 |
| | <<<Back 1 day (to 2020/06/26) | Forward 1 day (to 2020/06/28)>>> | |
Log of #mupdf at irc.freenode.net.