Commit 993e39a
committed
docs(plan): adopt two corrections from the parallel Windows investigation
bitkyc08-arch investigated the same CI red independently in
devlog/_plan/260731_windows_bun_stability/. The conclusions converge,
and two details there are more accurate than mine.
panic(thread is not a usable fingerprint: the thread token varies by
run (panic(thread 2852) in one, panic(main thread) in another). The
stable strings are 'oh no: Bun has crashed' and 'Internal assertion
failure'.
Timeouts do not appear on the (fail) line. They print separately as
'^ this test timed out after Nms', so grepping (fail) alone cannot tell
a budget overrun from a logic failure. Checking run 30618324009 with
the right pattern shows all three failures were budget overruns, and
executeArchivedCleanup blew a 20s budget -- the runner was that slow,
not the test that wrong.
That unit also sampled 10 failing runs to establish crash and
test-failure are mutually exclusive; I had looked at 2.
Keeping both documents. This one records what this round changed; that
one is the mode taxonomy and the candidate fixes.1 parent ee7c252 commit 993e39a
1 file changed
Lines changed: 32 additions & 0 deletions
Lines changed: 32 additions & 0 deletions
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
160 | 160 | | |
161 | 161 | | |
162 | 162 | | |
| 163 | + | |
| 164 | + | |
| 165 | + | |
| 166 | + | |
| 167 | + | |
| 168 | + | |
| 169 | + | |
| 170 | + | |
| 171 | + | |
| 172 | + | |
| 173 | + | |
| 174 | + | |
| 175 | + | |
| 176 | + | |
| 177 | + | |
| 178 | + | |
| 179 | + | |
| 180 | + | |
| 181 | + | |
| 182 | + | |
| 183 | + | |
| 184 | + | |
| 185 | + | |
| 186 | + | |
| 187 | + | |
| 188 | + | |
| 189 | + | |
| 190 | + | |
| 191 | + | |
| 192 | + | |
| 193 | + | |
| 194 | + | |
0 commit comments