| | <<<Back 1 day (to 2015/05/10) | 20150511 |
| aksr | hello. what happened to the mudraw -t(tt) option? | 04:19.11 |
| | i'm guessing it evolved into mudraw -F (txt, html...) | 04:32.16 |
| kens | chrisl here's a good one I spent the weekend chasing. I was getting a seg fault when discarding text, turns out that if we do a stringwidth operation, op_show_continue effectively doe a GRESTORE ON THE *ASSUMPTION* that the texet_begin method had done a gsave O.O | 08:12.57 |
| tor8 | aksr: you're correct. the -t(t(t)) option evolved into -F text,html,stext | 09:15.48 |
| | you can also just use the output filename, so mudraw -o out.txt will also work | 09:16.04 |
| | sebras, the first one is EPUB3 only. haven't seen the second one before, will take a look. | 09:18.10 |
| sebras | tor8: well, it would be nice if we just error out on the first one instead of crashing. | 09:37.23 |
| | tor8: despite not (yet?) supporting epub3. | 09:37.38 |
| | tor8: but I did try that testsuite and we never crash. | 09:37.46 |
| tor8 | we don't crash on epub3, we just print a warning that we might not support all features in the file | 09:38.17 |
| sebras | tor8: I know, I was just thinking if we see a strange css3 selector or something then the parser might do something weird. | 09:44.54 |
| tor8 | sebras: the parser will skip to the next '}' token if it encounters parsing errors and tries to continue from there | 09:45.43 |
| | and if it fails completely it'll print a message to the effect of 'ignoring stylesheet' | 09:46.20 |
| sebras | tor8: that is the intent, yes. but I wanted to see it in action. ;) | 09:46.49 |
| | tor8: I've also tried a bit of fuzzing but none has caused a crash. | 09:47.34 |
| tor8 | we support parsing the foo:n-th(even) type selectors now (we just 'forget' the (even) part though, since I don't save it in the fz_css_selectors yet | 09:47.47 |
| | apparently that was legal css2 syntax as well... | 09:47.58 |
| | sebras: just don't test for memory leaks when we run into errors... | 09:48.19 |
| sebras | tor8: ;) | 09:48.27 |
| chrisl | kens: I think I've stumbled across that one before..... | 10:14.05 |
| kens | I have too | 10:14.11 |
| | Caught me out again though | 10:14.16 |
| chrisl | Well, it's not an obvious thing | 10:14.40 |
| kens | Doesn't appear to be documented anywhere except in the op_show_restore procedure | 10:15.03 |
| | ANyway, I 'fixed' that, but somehow I've broken 2 other files in the process :-( | 10:15.28 |
| chrisl | I did look into it a while back, and I don't remember there being a good reason for it, either | 10:15.55 |
| kens | I can't see *any* reason for it, because it uses the saved level to restore back, so there shouldn't be any need to cater for any number of saves in the interim | 10:16.27 |
| | But.... I don't really want to try and debug it right now..... | 10:16.42 |
| chrisl | Likewise.... | 10:17.20 |
| kens | Oh damn, I think I see my problem. The stringwidth code pushes the null device. | 10:18.17 |
| | Robin_Watts : ping | 10:59.15 |
| Robin_Watts | pong | 10:59.22 |
| kens | I was reading a bit about bicubic downsampling over the weekend | 10:59.35 |
| | THe consensus seems to be that there is no single 'bicubic downsampling' algorithm | 10:59.50 |
| | Ands teh Mitchell filter is one of those quoted as a 'bicubic' downsamplign filter | 11:00.12 |
| Robin_Watts | That's fair. | 11:00.28 |
| kens | So is there any reason for us not to alwyas use the Mitchell filter in pdfwreite when downsampling and 'bicubic' is selected ? | 11:00.34 |
| Robin_Watts | 'Bicubic interpolation' is a well understood thing; see: http://en.wikipedia.org/wiki/Bicubic_interpolation | 11:00.51 |
| kens | Yeah, I was lookgin specifically for downsamplign though | 11:01.11 |
| Robin_Watts | But for downsampling it doesn't work once you get beyond a power of 2. | 11:01.13 |
| | Using the mitchell filter for downscaling would certainly be a reasonable approach. | 11:01.38 |
| kens | THen I thnk I'll do that. | 11:01.45 |
| | Other approaches with pdfwrie would be tricky | 11:01.57 |
| Robin_Watts | I might be tempted to experiment with the idea of only using the mitchell filter once the downscale is small than a factor of 2. | 11:02.03 |
| kens | That's a possible one, its about the only one that is possible mind you:-) | 11:02.19 |
| Robin_Watts | (I think that's doable in pdfwrite) | 11:02.24 |
| | yeah. | 11:02.26 |
| kens | When I get back to the poitn where I can play with master again I'll try out both | 11:02.56 |
| | I'll reassign that bug to me. | 11:03.04 |
| Robin_Watts | ok, thanks. | 11:03.31 |
| kens | Thanks for lookign at it :-) | 11:03.41 |
| Robin_Watts | np. | 11:04.56 |
| kens | Sooner or later I'm going to have to stop ignoring bugs...... | 11:05.08 |
| aksr | tor8: ok, thank you | 12:07.08 |
| kens | Hmm, I'm seeing 2 regression files which keep on producing an 'Internal_error' fikles are Bug695733.ps and Bug695860.ps. Anyone know if these are new files ? Or any idea why they are thrwoing an internal error ? | 12:30.01 |
| Robin_Watts | Bug 695733 is one of yours :) | 12:43.52 |
| kens | Yes, but I didn't put the file into the test suite | 12:55.21 |
| | Neitgher file actually produces any output, since they lack a showpage, and I wonder if ths is the problem | 12:55.51 |
| | Bug 695860 is one of mine also | 12:56.20 |
| | I tried pushign the current HEAD of origin/master and it behaves the same, so I'm assuming its not something I've broken..... | 12:57.51 |
| henrys | kens: so there are cases where text beging doesn't gsave and it isn't an error? | 13:34.48 |
| | s/beging/begin | 13:35.00 |
| kens | Its an 'implcit' gsave for dstringwidth | 13:35.08 |
| | In gx_default_text_begin I believe | 13:35.54 |
| henrys | I see | 13:37.06 |
| kens | The code in op_show_restore checks the 'save level' stored on the execution stack against the save level of the current gstate and restores back as many times as needed to make them the same. But in op_show_restore, before it does that, if its a strng width it reduces the saved level by 1 because there's a gsave in gx_default_text_begin for a stringwidth operaton. | 13:37.55 |
| | If your code handles stringwidth, then you need to know that and cater for it, or you end up doing the wrong numebr of grestores and crashing | 13:38.19 |
| | And I'm not convinced its required. If it is, there ought to be a better way of dealing with it than this kind of black magic | 13:38.45 |
| | Or at the very least it should be clearly documented. | 13:38.57 |
| | On the other hand, I could say the same about pretty much all the device methods. | 13:39.11 |
| henrys | yeah I was going to say... it's consistent ;-) | 13:39.36 |
| kens | Hmm, so now the 2 files causing me problems are no longer appearing in my cluster report, and I see marcos running a cluster test, I wonder if these events are connected. | 15:27.48 |
| Robin_Watts | oooh. | 15:47.09 |
| | Visual Studio community edition "An unlimited number of users within an organization can use Visual Studio Community for the following scenarios: in a classroom learning environment, for academic research, or for contributing to open source projects." | 15:47.32 |
| kens | Hey that's us :-) | 15:47.44 |
| Robin_Watts | So we can all use VS2015 for GS and MuPDF. But not for SO. | 15:47.47 |
| henrys | SO development does contribute to open source projects ;-) | 15:59.20 |
| Robin_Watts | I think it'd be fairer to say at the moment that open source projects contribute to SO, but... :) | 16:01.12 |
| mvrhel_laptop | wtf. how could my one little change in the transparency code cause all the diffs in the cluster testing of PS files | 17:28.14 |
| | oh they are pdfwrite | 17:28.39 |
| rayjj | does the VS 2015 Community Edition include a profiler that actually works ? | 17:29.45 |
| | It's hard to tell because it lists "Performance and Diagnostics Hub" as included, but not "IntelliTrace Performance Indicators" (whatever that is) that is part of the Enterprise edition | 17:32.22 |
| | mvrhel_laptop: did you update your repo before cluster testing. Somebody had a change recently (chrisl ?) that changed the text on every file, and I tripped over that | 17:34.25 |
| | hmm... it looks like the Performance and Diagnostics hub does have both CPU sampling and Instrumentation profiling | 17:39.22 |
| mvrhel_laptop | Robin_Watts: thanks for sending that email | 18:09.56 |
| Robin_Watts | mvrhel_laptop: No worries. | 18:10.05 |
| | mvrhel_laptop: So do we need to check with Miles exactly what they've been licensed? | 18:10.20 |
| | Do they get the entire gsview download? | 18:10.31 |
| mvrhel_laptop | Robin_Watts: we should probably check but I believe it is just the mupdfnet project part | 18:11.08 |
| | I will check with him | 18:11.36 |
| Robin_Watts | mvrhel_laptop: I suspect that Miles will have licensed the whole thing. | 18:11.41 |
| mvrhel_laptop | Maybe | 18:11.53 |
| Robin_Watts | I was expecting Miles to come back to us and say "what are they actually asking for?" :) | 18:12.04 |
| mvrhel_laptop | I did give him a heads up about this last week | 18:14.00 |
| | just sent him an email | 18:14.07 |
| | Robin_Watts: ok so just the .net project so I will get this stuff reorg to make that easier. | 18:29.46 |
| Robin_Watts | mvrhel_laptop: Great. | 18:29.54 |
| | Forward 1 day (to 2015/05/12)>>> | |
IRC Logs