| <<<Back 1 day (to 2018/07/03) | 20180704 |
paulgardiner | tor8: did you get a chance to look at signature-field appearance streams? I think once that's in, there's nothing stopping me taking on all your new stuff in muso. | 08:58.33 |
tor8 | paulgardiner: sorry, was distracted chasing down your other issues | 09:07.29 |
| I still hope to have that done by friday | 09:07.35 |
| paulgardiner: btw, I did waste some time rejigging the matrix and rect API for a proposal that's on tor/nostar | 09:08.11 |
paulgardiner | Oh, that would be great | 09:08.13 |
tor8 | if that goes in, it'll cause you a brief moment of pain while you change things from passing pointers to values (but brief and quickly fixed, and hopefully the end result will be cleaner code and worth it) | 09:09.15 |
paulgardiner | I prefer the passing by value style: far more readable, but didn't Robin perform some timing tests that showed it did actually cost us significantly? | 09:11.19 |
tor8 | paulgardiner: I redid some timing tests and find vague conclusions that passing by value might possibly end up being faster, but nothing conclusive can be measured | 09:12.05 |
paulgardiner | Okay then. Sounds good to me. | 09:12.24 |
Robin_Watts | tor8: http://git.ghostscript.com/?p=user/robin/mupdf.git;a=commitdiff;h=797f898669d69bf0e5a0c75a3c03a07296c12556 | 09:13.32 |
tor8 | Robin_Watts: change is_indirect to pdf_is_stream perhaps? | 09:14.34 |
Robin_Watts | tor8: yeah, I wondered. I could live with that. | 09:40.19 |
tor8 | Robin_Watts: alternatively, only "if (!pdf_name_eq(... Identity)" and error on anything that is neither a stream (because we try to open it) or is not Identity | 09:41.21 |
Robin_Watts | yeah. if it's a stream load it. if it's a name, and it's identity, silently do nothing, otherwise if it exists, warn. | 09:41.59 |
| warn rather than error, probably, cos error stops anything displaying. | 09:42.14 |
tor8 | Robin_Watts: or if (pdf_is_stream()) ... else if (not Identity) warn | 09:42.17 |
Robin_Watts | that. | 09:42.26 |
tor8 | go for it. | 09:42.30 |
Robin_Watts | I'll put that together. | 09:42.32 |
| tor8: http://git.ghostscript.com/?p=user/robin/mupdf.git;a=commitdiff;h=92668c4346480ff4038ecd9ee7beacb688900392 | 09:57.36 |
tor8 | Robin_Watts: LGTM | 10:12.07 |
Robin_Watts | Ta. | 10:12.30 |
tor8 | paulgardiner: on top of tor/master (if you can cope with the pass-by-value changes) is a commit for digital signature appearances | 10:36.18 |
paulgardiner | tor8: great ta | 10:36.41 |
tor8 | I had to fake a test button in mupdf-gl because I don't know how to hook up actual signing. I should probably look into that :) | 10:36.43 |
paulgardiner | I believe we have signature verification at least in the DebugOpenssl Windows build, and I have a commit that adds signing to that. | 10:38.51 |
tor8 | I find nothing in the source that calls pdf_sign_signature | 10:39.25 |
paulgardiner | yeah, that's in my commit | 10:39.45 |
tor8 | only pdf_check_signature | 10:39.47 |
| ah! | 10:39.49 |
| right, sorry, misunderstood you. | 10:39.55 |
paulgardiner | tor8: it doesn't look quite right, but it will probably do for now. The righthand side looks not to have used as much as the vertical space as it could have, plus there's supposed to be a symbol. | 11:06.30 |
| What's the preferred way to share an image on IRC? | 11:07.24 |
tor8 | paulgardiner: yeah, I didn't put in the logo and I just made sure the right hand side should fit within 6 lines | 11:08.19 |
paulgardiner | Ah okay. That matches what I see | 11:08.56 |
tor8 | It looks similar to Adobe's appearance apart from not vertically centering the right-most text | 11:09.22 |
| I think it should be good enough for now at least. | 11:09.27 |
paulgardiner | I've just pushed my windows signing commit to my repo in case it's useful, branch windows-signing. Runs off your current master | 11:13.14 |
tor8 | paulgardiner: thanks. | 11:19.42 |
| that looks simple enough | 11:20.07 |
| (if you have a pfx file around...) | 11:20.20 |
paulgardiner | In the post | 11:22.20 |
tor8 | paulgardiner: ah, your commit is missing winquery and wingetcertpath | 11:27.20 |
| but maybe those only exist on windows, not on x11 | 11:27.30 |
| I'll see about porting this to mupdf-gl | 11:28.30 |
paulgardiner | I think that's right. Doesn't look like I missed anything from the commit. | 11:29.00 |
tor8 | paulgardiner: no, it's just missing the stubs for x11 | 11:29.12 |
| paulgardiner: I'm getting segfaults in drop_signer when signing fails and I try to clean up. | 15:09.46 |
| paulgardiner: if you give a wrong password to the certificate it crashes | 15:10.24 |
paulgardiner | Oh right. I'll take a look | 15:10.44 |
| tor8: fix on that same branch | 15:19.26 |
| Along with a fix for another problem I found earlier. | 15:19.43 |
| At the moment, I'm trying to figure out what else we should be checking as part of verification. There could be a new incremental update since the last signing. We don't currently report that. I'm not sure how to handle it. | 15:21.17 |
tor8 | paulgardiner: thanks, that fix worked. | 15:26.12 |
| paulgardiner: I still get verification errors on the sample file, "Self-signed certificate" | 15:26.31 |
| is there a way to know that the file matches the certificate even if the certificate cannot be verified itself | 15:27.08 |
paulgardiner | The way things are at the moment, that is what that message means. If the file has been changed, that will be reported instead. | 15:28.05 |
tor8 | paulgardiner: right. have you got a file that goes through verification completely? | 15:28.51 |
paulgardiner | tor8: I do. Just the one file, but that one now seems to fail on loading, in a way unrelated to signatures. | 15:37.42 |
| An fz_image_key with an image 0xcccccccc | 15:39.14 |
| Ah no, it looks like *ctm where ctm = NULL is the problem | 15:40.29 |
| Could be the matrix/rect change | 15:41.10 |
| This is in fz_get_pixmap_from_image, where it calls fz_compute_image_key, maybe it wants ctm ? *ctm : id_transform | 15:43.44 |
| I mean ctm ? *ctm : fz_identity | 15:46.00 |
| Looks like that works, and now the file loads and the signature checks okay. I'll send you the file | 15:46.24 |
tor8 | paulgardiner: could be, toss me the file and I'll have a look. | 16:40.35 |
fredross-perry | tor8, paulgardiner: I've got a list popup working when the user taps on a form field that is a list or combo box. Is there something in the file format, and in the API, that represents what an author might have provided as a prompt? | 16:58.04 |
theatomheart | when i click on a hyperlink in mupdf it opens in an unpreferred browser and i can't seem to find where/how to change that to use $BROWSER | 17:27.23 |
paulgardiner | fredross-perry: I don't think there is. I believe the design intends in-place filling of forms, which we should probably work towards in time. | 17:32.31 |
tor8 | fredross-perry: paulgardiner: I think maybe the /TU entry might be something to that effect? | 17:34.47 |
fredross-perry | hmmmk. | 17:35.01 |
tor8 | nothing in the spec, but adobe seems to have it | 17:35.38 |
fredross-perry | I guess an in-place pop-next-to would be fine. | 17:35.57 |
tor8 | or the /T entry (field name) | 17:36.30 |
| see 8.6.2 Field Names in the spec | 17:36.57 |
fredross-perry | So, I am looking at Adobe Fille & Sign as an example. | 17:36.58 |
| as a UI example. | 17:37.07 |
| they are not even doing lists and combox properly. | 17:37.31 |
| that I can see. | 17:41.43 |
| Forward 1 day (to 2018/07/05)>>> | |