| | <<<Back 1 day (to 2018/06/13) | 20180614 |
| tor8 | paulgardiner: I'm not convinced looking at your code will solve the problem | 10:57.40 |
| | I just compared your sample file with what the old code synthesized for a form field: https://ghostscript.com/~tor/form-filling.png | 10:57.58 |
| | the "VENDOR NAME" string. on the left is the original file contents, on the right is what mupdf 1.12 synthesizes if the AP is missing. | 10:58.33 |
| | and editing the same field in acrobat also shifts the text position rather drastically | 11:01.25 |
| paulgardiner | Is the "VENDER NAME" label a form field? | 11:27.45 |
| | What does AR do with the version of that file where the AP is removed? | 11:29.57 |
| | That's maybe not the best test. Comparing with the placement that AR gives when form fields are filled out might be better. | 11:31.06 |
| tor8 | paulgardiner: it is a form field | 11:33.38 |
| | it approximates the horizontal alignment of the original, but the vertical alignment is more in line with your code | 11:34.06 |
| paulgardiner | That fits with my memory. I didn't think my code was correct, but it was closish. | 11:34.52 |
| tor8 | the 'new' code on tor/ui matches the vertical alignment of the original better | 11:35.16 |
| | bxut it's not what adobe does | 11:35.19 |
| paulgardiner | Yeah, but in another file I have here, it's looking quite far out from AR when filling form fields. | 11:36.43 |
| | I'd have thought a match to AR is what we want, because we are more usually trying to match what AR does when a form is being filled out than dealing with a missing appearance stream. | 11:38.00 |
| | Are those form fields that are marked as not editable then? I have no recollection of such things. | 11:38.54 |
| tor8 | paulgardiner: only one of the 'labels' are form fields | 11:40.14 |
| | the VENDOR NAME one | 11:40.22 |
| | and it's got a much higher object number, so I suspect it's been edited in after the rest of the form was created | 11:40.39 |
| | the other fields have object numbers 20-40 or thereabouts, but the VENDOR NAME one has number 394 | 11:40.56 |
| paulgardiner | I've sent you another file. Admittedly, the fields were filled out with the old mupdf, so you aren't looking to avoid movement with respect to them completely, but the placements aren't far off AR. Currently they jump a lot with the ui branch code. | 11:43.29 |
| tor8 | paulgardiner: yeah. I've tweaked the positioning mostly lookign at FreeText annotations | 11:44.04 |
| | I suspect the behaviour for padding and borders needs to be rather different for Widget/Tx annotations | 11:44.25 |
| paulgardiner | That could be it | 11:44.33 |
| tor8 | but thanks for the file. I'll tweak till it matches. | 11:47.15 |
| | paulgardiner: yeah, I see the jumps. | 11:51.06 |
| | is it 'centering' it vertically in a single-line form? what if you toggle it to multiline? | 11:51.23 |
| | the Requestx.pdf file matches quite well what adobe does, at least | 11:52.09 |
| paulgardiner | Oh, I don't know. But yes: makes sense that multiline would layout differently | 11:52.55 |
| | I have some recollection of alignment flags in PDF too. | 11:53.33 |
| tor8 | paulgardiner: page 2 on that form doesn't seem to format the fields correctly (the 0% fields) | 12:06.13 |
| | if I fill in a 0 it formats as blank | 12:06.20 |
| | and when I edit it, it shows up with the '%' in the value to edit, which acrobat doesn't | 12:07.04 |
| | is the 'format' constructor only to be invoked when generating the appearance, and not modify the value itself perhaps? | 12:07.37 |
| | s/constructor/callback/ | 12:07.45 |
| | and in mupdf 1.12 if I edit the form fields on page 2 in mupdf-win32 the changes don't show up on the page | 12:09.27 |
| | but on page 1 they do | 12:09.30 |
| | paulgardiner: hm, your code doesn't handle blank lines either... :) | 12:11.30 |
| paulgardiner | Oops! | 12:11.55 |
| Robin_Watts | reboot. | 12:18.47 |
| | .quit | 12:18.48 |
| avih | tor8: it seems it's impossible to disable readline in mujs if uname returns "Linux", e.g. while cross building for windows. i tried adding -UHAVE_READLINE to XCFLAGS but then it still tried to link -lreadline, which mingw doesn't have | 20:56.00 |
| | (also, wouldn't you want readline for osx? bsd?) | 20:56.41 |
| | you don't have configure for mujs, so imo readline should be disabled by default | 20:58.30 |
| | Forward 1 day (to 2018/06/15)>>> | |
Log of #mupdf at irc.freenode.net.
