Low Level Tools/2018/09/24
From MozillaWiki
Status updates for the week of 2018/09/24
Nick
rustc landings:
- Merge `bitvec.rs` and `indexed_set.rs`. A big clean-up, and prerequisite for...
- Use `HybridBitSet` in `SparseBitMatrix`. Reduced NLL peak memory by 14--45% on four rustc-perf benchmarks, and 96% on one external crate!
- Split `Liveness::users` into three. Reduces non-NLL peak memory by 20% on keccak.
- Compress `Liveness` data some more. Reduces instruction counts on NLL benchmarks by up to 16%, and max-rss by up to 38%.
- Improve handling of type bounds in `bit_set.rs`. A code clean-up.
Gabriele
- I triaged September 15th's nightlies and filed bug 1491761 - Crash in js::CloneRegExpObject
- I reviewed bug 1388134 which lazifies reading application directories in the crash manager
- I've finished and landed the patch for bug 1463048 which strips out asynchronous minidump generation as it proved unstable and hard to fix
- I closed four bugs that had become invalid or were fixed by the removal of the disk space wather: bug 1328321, bug 1206074, bug 876850 and bug 863596
- I've landed bug 1490240 which removes all the remaining warnings from the crash reporter code (and some from mozglue too) when building under Windows with clang-cl, unfortunately the patch was backed out for breaking some mozglue code
- I've filed and fixed bug 1492815 which ensures that Fennec and GeckoView do not send the 'StackTraces' field when submitting a crash report. This field is large and redundant as it's being used only for generating crash pings
- I've investigated some intermittent failures in bug 1376773 which tests crash detection in marionette but couldn't fix the issue yet
- I landed a fix for bug 1470223 which will improve the quality of the stack-traces printed out in the logs for try runs
- I've prepared a fix for bug 1493365 which was a gtest failure triggered by bug 1470223. It was a hard one to figure out because it was indirectly caused by the increase in memory consumption caused by the latter