While W3 (#10) was being built, one run of test_streaming_across_the_island_stays_under_the_frame_budget recorded a single 238 ms main-thread frame in TerrainView._process. The host's load average was about 10 at the time. The next 7 runs measured 0.8 to 1.4 ms worst-case, and the test passes.
The likely cause is the main thread being descheduled while worker threads and other host processes compete for CPU, but that isn't confirmed.
Acceptance
If the test ever fails on this again, record which part of the frame took the time (uploads, queueing or task reaping) before changing anything.
During H1, profile streaming on a real GPU and check whether WorkerThreadPool low-priority tasks starve the main thread on machines with few cores.
Close this if H1 and H4 show no spikes.
**Depends on:** W3
While W3 (#10) was being built, one run of `test_streaming_across_the_island_stays_under_the_frame_budget` recorded a single 238 ms main-thread frame in `TerrainView._process`. The host's load average was about 10 at the time. The next 7 runs measured 0.8 to 1.4 ms worst-case, and the test passes.
The likely cause is the main thread being descheduled while worker threads and other host processes compete for CPU, but that isn't confirmed.
**Acceptance**
- If the test ever fails on this again, record which part of the frame took the time (uploads, queueing or task reaping) before changing anything.
- During H1, profile streaming on a real GPU and check whether `WorkerThreadPool` low-priority tasks starve the main thread on machines with few cores.
- Close this if H1 and H4 show no spikes.
claude
added this to the v0.9 Hardening milestone 2026-11-10 19:28:54 +00:00
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Depends on: W3
While W3 (#10) was being built, one run of
test_streaming_across_the_island_stays_under_the_frame_budgetrecorded a single 238 ms main-thread frame inTerrainView._process. The host's load average was about 10 at the time. The next 7 runs measured 0.8 to 1.4 ms worst-case, and the test passes.The likely cause is the main thread being descheduled while worker threads and other host processes compete for CPU, but that isn't confirmed.
Acceptance
WorkerThreadPoollow-priority tasks starve the main thread on machines with few cores.Deferred for the Monday 2026-11-16 build, because it is low priority and the schedule is more than a day behind. This stays open. — tidewall-build