Perf: bump parallel workers 4 -> 6 (injection freed the CPU)
Android UI Tests / ui-tests (push) Has been cancelled

Pre-injection, 6 workers had occasional UI timeouts (the UI login was
CPU-heavy). With the login skipped via token injection, the machine has
headroom: a 6-worker run of the four heaviest suites passed 59/0/1 (was 1
failure pre-injection). 8 still thrashes. ~33% more parallelism, no coverage
cost.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This commit is contained in:
Trey T
2026-06-06 02:36:10 -05:00
parent d11cc82fec
commit a3b684744b
+4 -4
View File
@@ -34,10 +34,10 @@ SCHEME="HoneyDueUITests"
API_SCHEME="HoneyDueAPITests" API_SCHEME="HoneyDueAPITests"
TARGET="HoneyDueUITests" TARGET="HoneyDueUITests"
DESTINATION="platform=iOS Simulator,name=iPhone 17 Pro" DESTINATION="platform=iOS Simulator,name=iPhone 17 Pro"
# Data isolation removed cross-suite races, but each test now relaunches the # Workers. Bottleneck is CPU/simulator (each test relaunches the app). Token
# app + creates an account, so the bottleneck is CPU/simulator. 4 is the reliable # injection removed the CPU-heavy UI login, so 6 is now reliable (validated: 0
# sweet spot on a Mac mini (8 thrashed the simulators -> UI timeouts). Override via arg 2. # failures on the heaviest suites; 8 still thrashes). Override via arg 2.
WORKERS=4 WORKERS=6
# Suites that run in their own phases — excluded from the parallel phase. # Suites that run in their own phases — excluded from the parallel phase.
PHASE_MANAGED=( PHASE_MANAGED=(