Commit ee7c252
committed
docs(plan): finish the flake taxonomy — four classes, two opposite fixes
Chasing this to the end turned one problem into four, and two of them
take opposite treatment:
Bun runtime crash upstream, not ours
external binary in test removed the dependency (sidebar/gh)
wasted setup in a test merged four harnesses into one
budget below real work raised it (tray, claude ping)
The test for which treatment applies is whether the wait is intrinsic
to the assertion. Sidebar spawned gh to prove route reachability, so
the spawn was incidental and got deleted. The tray test launches
PowerShell, then Bun, then rebinds the port to prove the child never
inherited the socket -- the processes are the proof, so it gets a
budget instead.
The tray number came from arithmetic, not taste: production allows
PowerShell 15s while the test allowed 5s, and a contended runner falls
between. Observed failure was 5159ms.
Also records the condition that keeps this from reading as permission
to silence flakes: raising a budget requires the ablation to still
fail. Otherwise it is just silence, which is the fault 020 criticizes.1 parent 7093459 commit ee7c252
1 file changed
Lines changed: 38 additions & 0 deletions
Lines changed: 38 additions & 0 deletions
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
122 | 122 | | |
123 | 123 | | |
124 | 124 | | |
| 125 | + | |
| 126 | + | |
| 127 | + | |
| 128 | + | |
| 129 | + | |
| 130 | + | |
| 131 | + | |
| 132 | + | |
| 133 | + | |
| 134 | + | |
| 135 | + | |
| 136 | + | |
| 137 | + | |
| 138 | + | |
| 139 | + | |
| 140 | + | |
| 141 | + | |
| 142 | + | |
| 143 | + | |
| 144 | + | |
| 145 | + | |
| 146 | + | |
| 147 | + | |
| 148 | + | |
| 149 | + | |
| 150 | + | |
| 151 | + | |
| 152 | + | |
| 153 | + | |
| 154 | + | |
| 155 | + | |
| 156 | + | |
| 157 | + | |
| 158 | + | |
| 159 | + | |
| 160 | + | |
| 161 | + | |
| 162 | + | |
0 commit comments