| | <<<Back 1 day (to 2018/07/17) | 20180718 |
| chrisl_laptop | kens: I think the logic in cos_write_stream_alloc() is wrong..... | 16:57.54 |
| kens | really ? whereabouts ? | 16:58.10 |
| | Ah you mean the parent stuff ? | 16:58.37 |
| chrisl_laptop | gdevpdfo.c line 2048 | 16:58.51 |
| | Well, "wrong" may not be the best way to put it.... | 16:59.09 |
| kens | Err that's gs_md5_innit for me | 16:59.14 |
| chrisl_laptop | Well, about 5 lines below that: we save the device to the stream state | 17:00.03 |
| kens | yes | 17:00.08 |
| chrisl_laptop | It's not pdfwrite | 17:00.33 |
| kens | Correct, its not | 17:00.41 |
| | That's the point | 17:00.44 |
| | When we extract and use it later, we descend teh chain until we get to the end | 17:01.02 |
| NancyDurgin | wrong chat? | 17:01.37 |
| kens | There's a genuine problem with this whole area, because hte device when we come back to it, may not be the device when we stored it | 17:01.46 |
| | THat was an earlier fix | 17:01.52 |
| chrisl_laptop | NancyDurgin: No just didn't want to contaminate | 17:01.54 |
| kens | But you are correct that this is the cause of the seg fault Nancy is seeing with memento | 17:02.19 |
| | Fix finishing clustering now | 17:02.31 |
| chrisl_laptop | kens: I was concerned that elsewhere it might assume the stored device was pdfwrite, and write into it | 17:02.35 |
| kens | May not be the same problem as the non-Memento one though | 17:02.47 |
| | chrisl_laptop : it assuimes its pdfwrite and tries to *read* from it with disastrous consequences | 17:03.05 |
| chrisl_laptop | Ah, okay | 17:03.54 |
| kens | If the cluster test comes back clean I'll do a (slightly different) full test, I'm hopeful this may be the problem Nancy is chasing | 17:04.44 |
| chrisl_laptop | It would all be so much easier if we passed the gstate around, rather than the device :-( | 17:05.23 |
| kens | chrisl_laptop : yes quite true, in this case there's really no excuse for pdffwrite doing this, its just laziness on someone's part | 17:34.00 |
| kens | is being careful in case it was me | 17:34.07 |
| | The device *is* available at a higher level, but instead of changing the functions to pass it along, whoever it was just lazily copied the device into the structure. | 17:34.50 |
| | I may feel motivated to do soemthing more permanent about this when Nancy's stuff is working properly | 17:35.26 |
| | Forward 1 day (to 2018/07/19)>>> | |
Log of #ghostscript at irc.freenode.net.