Repository navigation
Commit 0e8af36
committed
test(S4b Zoo-Code-Org#1375): sort the acceptance row's four-name list and make its geometry assert itself
Two fixes in one file, both found by running things rather than by reading them.
The case-sensitive branch of the acceptance row asserted a lowercased entry list in
insertion order while the assertion itself sorts what the directory returns, so on a
case-sensitive filesystem the two orders differ and the row failed on the count
comparison even though every content assertion above it passed. The list is now in
sorted order, which is what the comparison asked for all along. The fact that the
content assertions ahead of it passed on ubuntu is the evidence that the rename-target
reading in the previous commit was right: the writer's own spelling carries this
writer's content, and the row's earlier failure there was the row asking the wrong file.
The acceptance row now also asserts its own geometry before it asserts anything about
the guard: the key handed to acquireFileLock has to be the spelling the writer asked in
with (the near half of the divergence), and the writer's own spelling has to exist by the
time acquisition returns (the far half). Both are recorded from inside the acquisition
seam, not inferred from the fixture.
That self-check earned its place immediately. Installed into the propagation row as
well, it came back with an empty capture array on this host: that row's doMocked
../fileLock seam never runs here, so the pair the row says it builds during acquisition
is not being built by that seam, and the stat count the row reports comes from the
folding filesystem's own restating rather than from the fixture. A diagnostic that says
"your double is not running" is not a verdict about the guard, so the capture and its
assertions were removed from that row and the stat-count assertion keeps a message that
says what a zero means. The propagation row's case-sensitive half therefore stays
unverified, and the question of which seam should supply the peer's appearance during
acquisition is open - it is a design choice about the lock machinery, not a line I land
on my own.
Local at this head: safeWriteJson.lockKey.spec.ts 10 passed (the case-folding branch;
the case-sensitive branch cannot execute on a case-insensitive host - a hard link onto
the other-case name answers EEXIST and readdir still lists one entry, measured). eslint
--prune-suppressions --max-warnings=0 exits 0 with src/eslint-suppressions.json back
equal to HEAD, prettier --check clean, tsc reports only the known
activate/registerCommands.ts(88,102) TS2739 junction artefact. Test-only, no platform
gate added.1 parent 258c5fe commit 0e8af36
1 file changed
Lines changed: 29 additions & 3 deletions
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
183 | 183 | | |
184 | 184 | | |
185 | 185 | | |
| 186 | + | |
| 187 | + | |
| 188 | + | |
| 189 | + | |
| 190 | + | |
| 191 | + | |
186 | 192 | | |
187 | 193 | | |
188 | 194 | | |
| |||
236 | 242 | | |
237 | 243 | | |
238 | 244 | | |
| 245 | + | |
| 246 | + | |
| 247 | + | |
| 248 | + | |
| 249 | + | |
| 250 | + | |
| 251 | + | |
239 | 252 | | |
240 | 253 | | |
241 | 254 | | |
| |||
247 | 260 | | |
248 | 261 | | |
249 | 262 | | |
| 263 | + | |
| 264 | + | |
| 265 | + | |
| 266 | + | |
| 267 | + | |
| 268 | + | |
| 269 | + | |
| 270 | + | |
| 271 | + | |
| 272 | + | |
| 273 | + | |
250 | 274 | | |
251 | 275 | | |
252 | 276 | | |
| |||
264 | 288 | | |
265 | 289 | | |
266 | 290 | | |
267 | | - | |
| 291 | + | |
268 | 292 | | |
269 | 293 | | |
270 | 294 | | |
| |||
550 | 574 | | |
551 | 575 | | |
552 | 576 | | |
553 | | - | |
554 | 577 | | |
555 | 578 | | |
556 | 579 | | |
| |||
562 | 585 | | |
563 | 586 | | |
564 | 587 | | |
565 | | - | |
| 588 | + | |
| 589 | + | |
| 590 | + | |
| 591 | + | |
566 | 592 | | |
567 | 593 | | |
568 | 594 | | |
| |||
0 commit comments