| <<<Back 1 day (to 2013/03/31) | 2013/04/01 |
henrys | Robin_Watts: It's strange we don't have a weekday holiday for easter here. Easter is widely celebrated. | 13:37.47 |
Robin_Watts | henrys: We have 2 here; Good Friday and Easter Monday. | 13:38.32 |
henrys | oh I didn't you got Friday too. | 13:39.17 |
Robin_Watts | I don't know that any of us took it though :) | 13:39.33 |
| my memory squeezing job finished. | 14:38.26 |
| 3 SEGVs left. | 14:39.07 |
| ok, memory squeezing segv's all fixed. | 17:34.07 |
| For 1 file, to one device at one resolution :) | 17:34.30 |
mvrhel_laptop | Robin_Watts: great! | 17:48.25 |
Robin_Watts | mvrhel_laptop: Someone else can fix the leaks :) | 17:56.33 |
henrys | Robin_Watts: why not a suppression file instead of PACIFY_VALGRIND? | 18:13.33 |
| we can store the file in the release | 18:13.52 |
Robin_Watts | henrys: because if we ever get a bug which looks like an indeterminism we can rerun with PACIFY_VALGRIND. | 18:32.14 |
| if it goes away, then we can see that maybe one of our pacifications is actually required. | 18:32.28 |
| Also suppression files are inherently tied to valgrind, PACIFY_VALGRIND would actually work with whatever tool we choose to use. | 18:33.45 |
henrys | but no recompilation and well understood by outsiders seem stronger than the define advantages to me. I guess we could do both and have owners decide which mechanism they prefer. | 18:38.10 |
Robin_Watts | The worry I have with the suppression files, is that it's very easy to say "Oh, yes, I understand why that's the case, just add it to the suppression file" and to find that you've actually misunderstood. | 18:48.52 |
| and so you've blocked stuff that's genuinely a problem. | 18:49.15 |
| With PACIFY_VALGRIND we add code that really is guaranteed to make valgrind run happily - and we can see that it doesn't affect the output (hence we get a correctness test too). | 18:50.15 |
| How did Marcos figure out who to assign these valgrind bugs to? Does he have a dartboard? :) | 18:51.58 |
mvrhel_laptop | with picture of me on it judging from my bug list ;) | 18:55.49 |
| lunch time. bbiab | 18:59.11 |
henrys | I guess he didn't run pcl | 19:15.05 |
Robin_Watts | so, I have a valgrind problem here where it's finding an undefined value in a tile. | 20:07.37 |
| If I set the tile to all 0's beforehand I get rid of the valgrind error. | 20:08.07 |
| and I get the same results out whether I set the tile to 0, don't set the tile, or set the tile to 0xFF. | 20:08.31 |
| So this looks like a false positive. | 20:08.43 |
| I wonder if valgrind has problems with bit operations on undefined values... | 20:09.00 |
| Forward 1 day (to 2013/04/02)>>> | |