| <<<Back 1 day (to 2018/06/12) | 20180613 |
tor8 | I bet I know why updating text fields is slower ... I need to be smarter about caching/finding the font resource | 09:04.58 |
| paulgardiner: ^ | 09:05.02 |
| paulgardiner: could you try the tip of tor/ui and see if that solves the slowdown? | 09:33.58 |
| paulgardiner: and if that doesn't help, is the bottleneck fz_md5_update? | 09:35.54 |
paulgardiner | tor8: okay, will take a look. | 09:52.10 |
tor8 | I have a patch to solve any slowdown from repeated md5 checksumming in pdf_add_simple_font too | 10:00.10 |
| which is also on tor/ui | 10:00.16 |
paulgardiner | tor8: that looks to be loads better. By eye, I now don't see its being any slower than the version on master. | 10:18.19 |
tor8 | paulgardiner: fab! | 10:22.29 |
paulgardiner | I'm just going to retest on master the file from which I cannot delete annotations | 10:23.00 |
| Yeah, that happens on master too, although I have just discovered another problem on the ui branch: In landscaps files, contents of fields are being rotated by 90deg when altered | 10:29.35 |
| landscape files | 10:29.48 |
| But then again, you haven't really tackled appearance streams for form fields, so perhaps that's not surprising. | 10:30.53 |
tor8 | paulgardiner: hmm, there may be some generic problem with rotated pages and all annotations. | 10:33.13 |
| I suspect once I fix that, it'll fix all of them | 10:33.23 |
paulgardiner | Strangely it doesn't happen with text widgets on master, although in the same file a text annotation icon is rotated | 10:36.28 |
tor8 | paulgardiner: yeah... seems like the old code has the same problem with rotated pages with Text annotations at least. | 11:08.29 |
| oh dear... even acrobat seems to be weird with this. | 11:13.59 |
| I created a sticky note in acrobat on a rotated page | 11:14.01 |
| and when I draw that with mupdf (no synthesis) it doesn't match position with what acrobat has, nor with what evince does | 11:15.05 |
| so something's off even with plain annotations and no appearance synthesis | 11:15.19 |
| same behavior with both master and ui branches | 11:15.40 |
| oh, gs also draws it unrotated. that's a bit more comforting! | 11:30.17 |
| paulgardiner: what file do you have that has rotated pages with text fields? | 12:53.19 |
paulgardiner | I'll email it to you | 12:53.46 |
tor8 | my brain is melting in transform hell... | 12:54.22 |
paulgardiner | It's also the file which I can't delete annotations from | 12:54.29 |
tor8 | I think some of our problems are down to not respecting the NoRotate annotation flag | 12:54.49 |
| and trying to respect that just caused my brain to implode... | 12:55.12 |
paulgardiner | There are some text fields that are initially displayed correctly, but rotate when updated | 12:55.45 |
tor8 | paulgardiner: i can delete annotations in that file in mupdf-gl | 12:59.15 |
| paulgardiner: ah yes, the /MK dictionary that I don't currently use | 13:02.16 |
| and Adobe seems to also use an undocumented field /Rotate in its annotation dictionaries | 13:03.13 |
| a field that's not mentioned anywhere in any spec... | 13:03.26 |
kens | I seem to recall having to add support for that | 13:06.18 |
tor8 | kens: MK or NoRotate? | 13:08.14 |
| or annotation/Rotate? | 13:08.25 |
kens | Rotate in annotations | 13:08.28 |
| I don't think GS uses the other stuff | 13:08.45 |
tor8 | gs doesn't seem to respect the NoRotate flag at least | 13:09.06 |
kens | Well, there's a comment about it in pdf_draw.ps | 13:09.21 |
| Bit 5 NoRotate | 13:09.28 |
| But it look slike we don't honour Rotate in Annotations (I oculd be wrong, the PS code is hideous) | 13:10.29 |
tor8 | I don't think gs handles any of it... maybe Jaws did? | 13:10.53 |
kens | No Jaws didn't handle annotations at all really | 13:11.09 |
| I'm sure it does by now | 13:11.15 |
| I just had a memory of doing Rotate for annotations in GS | 13:11.31 |
| And it looks like the comment in the GS interpreter is exactly that, just a comment | 13:11.50 |
tor8 | and I only see gs handling /MK/BG and /MK/BC but not /MK/R | 13:14.29 |
kens | Yeah I know GS's annotation synthesis is far from complete | 13:14.52 |
| I add new stuff from time to time when it gets reported | 13:15.00 |
| But I start off from the premise of 'put in an appearance stream' | 13:15.17 |
tor8 | paulgardiner: rotated text field fix on tor/ui | 14:21.06 |
paulgardiner | Ah magic | 14:21.18 |
tor8 | still need to look more at the baseline calculation | 14:21.19 |
| paulgardiner: turned out to be trivial, once you remember how the 'magic' annot/BBox/Matrix/Rect interact. | 14:21.37 |
| paulgardiner: could you see if newest commit on tor/ui fixes the baseline jumping? | 14:31.19 |
paulgardiner | Yeah, sure. | 14:39.27 |
| Just preparing a muso commit. Will look once I've done that | 14:40.02 |
tor8 | paulgardiner: I doubt it will, but if it does it would be nice to know. | 14:40.41 |
paulgardiner | tor8: no: still jumps. As well as moving to the top of the box, it's moving a little to the right. I didn't notice that before, but possibly that's not new | 15:00.10 |
tor8 | the motion to the right is new | 15:06.09 |
| I'll study some more | 15:06.17 |
Robin_Watts | Morning mvrhel_laptop | 15:23.57 |
mvrhel_laptop | Morning Robin_Watts . I just sent you an email | 15:24.19 |
| I am off to the Dentist in a bit | 15:24.32 |
| Thanks for your excellent comments on the paper | 15:24.43 |
inflex | Is it possible to make muPDF-gl "flash" the screen or something to indicate no more search hits? I was thinking of just running "invert colours" once. | 15:45.42 |
CodyActivePDF | hello @Robin_Watts. I'm cloning a fresh copy of Mu to be absolutely sure that the issue isn't something I did while tinkering the last few days. I'll send you build logs once I have finished cloning. | 16:49.27 |
Robin_Watts | Hi CodyActivePDF | 16:52.41 |
| Just so you're aware, this is a public channel. | 16:52.57 |
| (i.e. don't say anything commercially sensitive here :) ) | 16:53.14 |
CodyActivePDF | @Robin_Watts the fresh copy of Mu I just cloned compiles just fine, so I must have screwed something up on the other build in the last week. Thanks for your help! | 17:11.28 |
Robin_Watts | CodyActivePDF: Ah, cool. | 17:11.39 |
| Sorry it took a few iterations to get there. | 17:11.53 |
| If you ever have problems again, coming here may be the quickest way to get help. | 17:12.16 |
| I'm generally online for working hours UK time (and a bit later). | 17:12.32 |
| My colleagues are in Sweden and Taiwan, so there will often be *someone* here :) | 17:12.50 |
CodyActivePDF | Understandable, software does that :) We're having a couple of issues with an older version of Mu, I'm going to attempt to reproduce them in this build and I'll follow with emails to the support email if I see the same issue. | 17:13.21 |
| Good to know, I'll bookmark this chat room. Thanks! | 17:13.32 |
Robin_Watts | CodyActivePDF: Cool. I'll watch out for that. | 17:13.43 |
tor8 | paulgardiner (for the logs): something possibly better with text alignment on tor/ui now | 17:44.50 |
| I don't think we can ever get it "perfect" ... text positioning in freetext and widget appearance streams is a bit of a voodoo problem. | 17:45.31 |
| Forward 1 day (to 2018/06/14)>>> | |