Repository navigation
Commit b994ce9
Run the tiny e2e capstones on 3 runners instead of 9 (#1734)
* Combine tiny e2e legs into three runners
Nine e2e matrix legs ran a few seconds of tests each on their own Depot
runner, so ~80% of each leg was runner start, checkout, binary download
and toolchain setup (#1732). Their toolchains don't overlap, so group
them:
- Poetry, PDM, Hatch and deno VEX capstones: one leg
- Pipenv and pip VEX capstones: one leg (same uv + Pythons setup)
- uv redirect, composer and .NET 10 capstones: one leg
A row's test_filter may now hold several ';'-separated filters. The run
step runs each suite once per filter in its own log group, and every
suite x filter must still run at least one test. The tier test compares
those expanded runs, so e2e-full rows still have to match a PR-tier run.
Fixes #1732
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SwAvjXtnjFQ1fNNf5f22eB
* Guard blank e2e filter segments and skip a duplicate uv install
A stray ';' in a combined test_filter produced a blank segment, which ran
the suite binary with no arguments: every non-ignored test ran, passed,
and the empty-run check could not tell. Fail the leg on a blank segment
instead.
The combined Poetry/PDM/Hatch/deno leg installed uv 0.11.19 twice: once
in the PDM/Hatch step and again in 'Install pinned Poetry'. Skip the
PDM/Hatch step when the leg also runs Poetry.
Found by /code-review high on #1734.
Co-Authored-By: Claude <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012WGa6qSVVc5KBepGwGx3cJ
* Skip blank e2e filters and tidy the matrix
Review follow-ups for the combined e2e legs:
- skip a blank `;` segment in test_filter instead of running the
binary with no arguments, which selects the non-ignored tests, and
make the tier test's runs() skip it the same way
- check the PDM row through runs() instead of a second parser
- put the PDM/Hatch, deno and .NET comments back next to the rows
they describe
- install uv 0.11.19 once for Poetry, PDM and Hatch instead of again
inside the Poetry step
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SwAvjXtnjFQ1fNNf5f22eB
* Run Pipenv in the short VEX leg, not with pip
The profiler's first runs of #1734 put the Pipenv + pip leg at
3.33-3.48 min, the last e2e job to finish in run 40411347506. pip
alone is p50 2.88 min and was already the last job before ci-ok in
2 of 7 merge_group runs on 10-26, so adding Pipenv's ~0.5 min there
lands on the merge-queue critical path.
Move pipenv:: into the Poetry/PDM/Hatch/deno leg (~1.1 min) and give
pip its own leg back. The runner count is unchanged (still 3 instead
of 9). In the shared leg the Pipenv setup-uv (0.13.5) is first on
PATH, so PDM/Hatch bootstrap with it instead of 0.11.19; that uv is
tooling there, not the release under test.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011f9aeP4Fqo6SktDCrA3ZuK
* Give Pipenv its own e2e leg
efa1296 moved pipenv:: into the Poetry/PDM/Hatch/deno leg. There the
Pipenv setup-uv (0.13.5) goes first on PATH, and PDM 2.29 bootstrapped
with it fails to lock: 'super(type, obj): obj (instance of
PDMPyPIClient) is not an instance or subtype of type (Client)'. Both
PDM capstones failed in run 40481294505.
Pipenv can't share that leg, and with pip it would add ~0.5 min to the
e2e critical path, so give it a leg of its own: 4 runners instead of
the original 9, and the shared leg goes back to the set that passed.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SwAvjXtnjFQ1fNNf5f22eB
* Revert "Give Pipenv its own e2e leg"
This reverts commit 452247e. Its premise was wrong: the PDM failure on
efa1296 was not caused by the Pipenv uv 0.13.5. On 452247e, with no
setup-uv in the leg, PDM failed the same way. The cause is unearth
0.19.0 (2026-10-27) moving to httpx2, which breaks `pdm lock` on every
PR; the next commit ports #1738's unearth<0.19 pin. Hatch passed
under uv 0.13.5, so the 3-runner layout from efa1296 stands.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SwAvjXtnjFQ1fNNf5f22eB
* Port #1738's unearth<0.19 PDM pin
unearth 0.19.0 (2026-10-27) moved to httpx2 and breaks `pdm lock` on
PDM 2.x, failing the pdm e2e leg (and ci-ok) on every PR. Carry the
same pin as #1738 so this PR can go green; it is a no-op once #1738
lands on main.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011f9aeSWjd3xeW8riRfmswG
---------
Co-authored-by: Claude <noreply@anthropic.com>1 parent 7e9b27a commit b994ce9
2 files changed
Lines changed: 76 additions & 38 deletions
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
1218 | 1218 | | |
1219 | 1219 | | |
1220 | 1220 | | |
1221 | | - | |
1222 | | - | |
| 1221 | + | |
| 1222 | + | |
| 1223 | + | |
| 1224 | + | |
| 1225 | + | |
| 1226 | + | |
| 1227 | + | |
| 1228 | + | |
| 1229 | + | |
| 1230 | + | |
1223 | 1231 | | |
1224 | 1232 | | |
1225 | 1233 | | |
| |||
1347 | 1355 | | |
1348 | 1356 | | |
1349 | 1357 | | |
1350 | | - | |
1351 | | - | |
| 1358 | + | |
| 1359 | + | |
1352 | 1360 | | |
1353 | 1361 | | |
1354 | 1362 | | |
1355 | 1363 | | |
1356 | 1364 | | |
1357 | | - | |
| 1365 | + | |
| 1366 | + | |
| 1367 | + | |
| 1368 | + | |
| 1369 | + | |
| 1370 | + | |
| 1371 | + | |
| 1372 | + | |
1358 | 1373 | | |
1359 | 1374 | | |
1360 | 1375 | | |
1361 | 1376 | | |
1362 | | - | |
1363 | 1377 | | |
1364 | 1378 | | |
1365 | 1379 | | |
1366 | 1380 | | |
1367 | | - | |
1368 | 1381 | | |
1369 | 1382 | | |
1370 | | - | |
| 1383 | + | |
| 1384 | + | |
| 1385 | + | |
1371 | 1386 | | |
1372 | 1387 | | |
1373 | 1388 | | |
1374 | 1389 | | |
1375 | | - | |
| 1390 | + | |
1376 | 1391 | | |
1377 | 1392 | | |
1378 | 1393 | | |
| |||
1407 | 1422 | | |
1408 | 1423 | | |
1409 | 1424 | | |
1410 | | - | |
1411 | | - | |
1412 | | - | |
1413 | | - | |
1414 | | - | |
1415 | | - | |
1416 | | - | |
1417 | 1425 | | |
1418 | 1426 | | |
1419 | 1427 | | |
| |||
1505 | 1513 | | |
1506 | 1514 | | |
1507 | 1515 | | |
1508 | | - | |
| 1516 | + | |
1509 | 1517 | | |
1510 | | - | |
| 1518 | + | |
| 1519 | + | |
1511 | 1520 | | |
1512 | 1521 | | |
1513 | 1522 | | |
| |||
1830 | 1839 | | |
1831 | 1840 | | |
1832 | 1841 | | |
1833 | | - | |
1834 | | - | |
1835 | | - | |
1836 | | - | |
1837 | | - | |
1838 | | - | |
| 1842 | + | |
| 1843 | + | |
| 1844 | + | |
| 1845 | + | |
| 1846 | + | |
| 1847 | + | |
| 1848 | + | |
| 1849 | + | |
1839 | 1850 | | |
1840 | 1851 | | |
1841 | 1852 | | |
1842 | 1853 | | |
1843 | 1854 | | |
1844 | 1855 | | |
| 1856 | + | |
1845 | 1857 | | |
| 1858 | + | |
1846 | 1859 | | |
1847 | 1860 | | |
1848 | 1861 | | |
1849 | 1862 | | |
1850 | 1863 | | |
1851 | | - | |
1852 | | - | |
1853 | | - | |
1854 | | - | |
1855 | | - | |
1856 | | - | |
1857 | | - | |
1858 | | - | |
1859 | | - | |
1860 | | - | |
1861 | | - | |
| 1864 | + | |
| 1865 | + | |
| 1866 | + | |
| 1867 | + | |
| 1868 | + | |
| 1869 | + | |
| 1870 | + | |
| 1871 | + | |
| 1872 | + | |
| 1873 | + | |
| 1874 | + | |
| 1875 | + | |
| 1876 | + | |
| 1877 | + | |
| 1878 | + | |
| 1879 | + | |
| 1880 | + | |
| 1881 | + | |
| 1882 | + | |
| 1883 | + | |
| 1884 | + | |
| 1885 | + | |
| 1886 | + | |
| 1887 | + | |
| 1888 | + | |
| 1889 | + | |
1862 | 1890 | | |
1863 | 1891 | | |
1864 | 1892 | | |
| |||
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
34 | 34 | | |
35 | 35 | | |
36 | 36 | | |
| 37 | + | |
| 38 | + | |
| 39 | + | |
| 40 | + | |
| 41 | + | |
| 42 | + | |
| 43 | + | |
| 44 | + | |
37 | 45 | | |
38 | 46 | | |
39 | 47 | | |
| |||
50 | 58 | | |
51 | 59 | | |
52 | 60 | | |
53 | | - | |
| 61 | + | |
| 62 | + | |
| 63 | + | |
54 | 64 | | |
55 | 65 | | |
56 | | - | |
| 66 | + | |
57 | 67 | | |
58 | 68 | | |
59 | 69 | | |
| |||
343 | 353 | | |
344 | 354 | | |
345 | 355 | | |
346 | | - | |
| 356 | + | |
347 | 357 | | |
348 | 358 | | |
349 | 359 | | |
| |||
0 commit comments