| | <<<Back 1 day (to 2012/08/03) | 2012/08/04 |
| halfie_ | Hi, I am looking for a sample Revision 5 encrypted PDF file. Any links? | 05:15.07 |
| | Acrobat 10 Pro won't generate revision 5 files. | 05:16.00 |
| sebras | tor8 Robin_Watts: I have rebased my patch at sebras/master ontopof origin/master now. it's one-liner... | 09:13.11 |
| Robin_Watts | Morning sebras. | 09:22.02 |
| sebras | Robin_Watts: morning mr. watts. | 09:22.14 |
| Robin_Watts | So, CSR (who merged with Peerless/Zoran a while ago), are being part bought by Samsung and relocated. | 09:22.43 |
| | http://www.csr.com/assets/common/Circular_1August2012.pdf | 09:22.52 |
| kens | Hmm, am I misreading that or is it just the phone business they are transferring Robin_Watts ? | 09:44.29 |
| Robin_Watts | yeah, I think so. | 09:44.47 |
| kens | Interesting though | 09:44.59 |
| sebras | morning tor8. | 11:37.42 |
| | when mupdf is repairing a pdf and happens to end up with a negative object number, what should we do? | 11:38.25 |
| | currently, we're trying to use the number as an index into xref->table, which seems silly. | 11:38.48 |
| | my approach is to set the object number to 0 under these circumstances. opinions? | 11:39.47 |
| | the same goes for object generation of course. | 11:40.03 |
| tor8 | sebras: so treat them as 'null' basically? | 11:49.05 |
| sebras | tor8: no, we map them to object number 0, the object which is supposedly always free. | 11:49.35 |
| | I haven't really checked but I think we can reference it. | 11:49.57 |
| | 0 0 R | 11:50.00 |
| | do you prefer just skipping those objects? | 11:50.16 |
| | I guess by the time a pdf is so broken that it has negative object numbers, then the likelihood of repair actually succeeding is pretty slim. this is on an artifically broken pdf, but passing it through mupdf makes it crash, naturally. | 11:51.22 |
| | also we never check whether the maximum object number encountered is reasonable. so if we encounter something ridiculous like 2**31 mupdf will try to resize the xref to that many object slots. | 12:02.02 |
| | tor8: ok, there's a patch over at sebras/master | 12:16.02 |
| | the stack overflow handling when lexing tokens in pdf_run_stream() is broken. I'm ending up in an eternal loop. | 12:38.31 |
| | tor8: interesting. I made fz_dash_lineto() take forever by generating points that are very far away from each other. this means that sqrt(dx^2 + dy^2) is long and that the dashing goes on forever. | 16:46.03 |
| | I mean, it is finite but it takes a very long time. | 16:46.23 |
| | Robin_Watts: why is there an assert in fz_free_context() for ctx->error->top == -1? | 17:10.19 |
| Robin_Watts | You pass the context pointer to fz_try() right? | 17:11.00 |
| sebras | yes. | 17:11.10 |
| Robin_Watts | so if you're freeing the context, you'd better not be in any fz_try's. | 17:11.18 |
| | hence the ctx->error->top = -1 | 17:11.27 |
| sebras | do I need to declare variables fz_var() if I'm setting them inside fz_catch()? | 17:12.30 |
| | I don't think so, but I'm not 100%. | 17:12.43 |
| Robin_Watts | You need to declare variables fz_var if you are setting them in the fz_try(). | 17:13.05 |
| | (basically, if you longjmp, you can lose variables being set) | 17:13.28 |
| sebras | and that is because they might be uninitialized if we enter fz_throw. | 17:13.30 |
| | right. | 17:13.34 |
| Robin_Watts | but if you're setting them in the fz_catch you should be fine. | 17:13.51 |
| sebras | Robin_Watts tor8: a few patches to review at sebras/master. | 17:37.37 |
| | at least once the weekend is over. :) | 17:37.46 |
| | hm... | 22:05.04 |
| | sane.sh doesn't work out of the box. it still refers to pdfdraw and xpsdraw. tor8 did you forget to push? | 22:05.26 |
| | tor8 Robin_Watts: yey, after messing about some with sane.sh I managed to run it. only log differences between origin/master and sebras/master. | 22:25.45 |
| | Forward 1 day (to 2012/08/05)>>> | |
IRC Logs