| <<<Back 1 day (to 2017/10/03) | 20171004 |
sebras | Robin_Watts: I'm confused. the cluster reports that my latest run differs from trunk when rendering fts_23_2311.pdf. on sebras/master it aborts, but it does so on origin/master too. I bisected a bit and discovered that commits prior to d0b78f4166a1503ce522944002b3aab035724cd9 renders the file without issues (but perhaps wrongly?) and after this commit we get a segfault | 13:12.53 |
| more recently, in commit dd58ea5e2ff7e6b61cc5a5b6fd950fb62e92b8be, the behaviour was changed into asserting instead (which is exactly what we still hav on origin/master). | 13:13.25 |
| the most puzzling thing is that the cluster reports a difference between origin/master and sebras/master, I don't get that. | 13:15.35 |
Robin_Watts | sebras: I don't know offhand. | 14:26.26 |
| sebras: different error code? :( | 14:26.41 |
sebras | Robin_Watts: nope, the log on sebras and on mupdf latest nightly both say return 0 for that file. | 14:27.51 |
| Robin_Watts: an both say "Command terminated by signal 11" (that's SIGSEGV) | 14:28.22 |
Robin_Watts | hmm. | 14:31.16 |
sebras | Robin_Watts: but when I run locally I see an assert being triggered. | 14:33.06 |
| Robin_Watts: I think that one might happen because the to_ccs/from_ccs function pointers are not set in fz_device_rgb() and friends. | 14:33.33 |
| and they are NULLed by fz_set_cmm_engine() when it replaces the default fz_device_rgb() which have to_ccs/from_ccs set with the ICC-based equivalents. | 14:34.10 |
| the assert is in std_conv_color: assert(srcs->to_ccs && dsts->from_ccs); | 14:34.55 |
| and it's attempting to convert from DeviceCMYK to DeviceRGB | 14:35.20 |
Robin_Watts | sebras: cluster builds release, so asserts don't trip. | 14:37.29 |
sebras | oh, that's why. | 14:38.10 |
| and then we he a segfault a few lines later. | 14:38.29 |
Robin_Watts | sebras: fz_var(chain) ? | 15:32.35 |
| sebras: I fear both your latest commits will need an fz_var(chain) added to them (just from looking at the diffs) | 15:42.47 |
sebras | Robin_Watts: true, "Make sure to drop chain upon error in raw and crypto filters." needs a fz_var(). | 15:46.49 |
| Robin_Watts: but I don't think "Drop stream upon error in inline stream." needs its. | 15:47.10 |
Robin_Watts | sebras: You are right. | 15:47.29 |
sebras | Robin_Watts: I ran these patches by tor8, but both he and I missed this. thanks for spotting this. | 15:48.09 |
Robin_Watts | sebras: No worries. It tickled my brain as being similar to a problem you spotted in an earlier commit. | 15:48.36 |
sebras | exactly, so I ought to have caught this myself. | 15:48.54 |
| Robin_Watts: patch on sebras/master | 15:51.03 |
Robin_Watts | lgtm. | 15:51.29 |
| Forward 1 day (to 2017/10/05)>>> | |