[JarryShaw]

• Searched for similar pull requests
• Followed the coding style (make pylint, make mypy, make isort)
• make test passes, and a test case covers the change
• Changelog entry: N/A — centralised in docs(changelog): shared 1.5.0 changelog — long-lived, merges last (#610, #616, #617, #618, #620) #657

What is the purpose of your pull request?

• ci — workflows or build tooling

---

Description of your pull request and other information

On main's GraalPy leg (job 118912680486), nine workers went down. Each one's Java heap grew across tests until it reached the --vm.Xmx2g cap (#1753). This PR runs the leg's tier as a series of pytest invocations, each with fresh workers.

The shard boundaries are chosen so that no shard's heap gets near the cap. A shard is a directory or a glob of one directory's entries, and every later run ignores what earlier runs took. A whole-directory split was not enough: tests/corekit alone reached 2028 MB.

Measured locally on GraalPy 25.4.4 (-n 3, Xmx2g):

	Runtime	Workers down	Peak live heap
Whole tier	3559 s	8	2038 MB
Sharded	1879 s	0	1303 MB

How it runs. util/pytest_shards.py runs inside the step's existing timeout, prlimit, scope and watcher, so those caps still cover the whole tier. On SIGINT it lets the current run write its JUnit file and starts no further runs. Each shard writes its own junit-<shard>.xml.

Unchanged:

• the heap flags, -n, the summaries and continue-on-error;
• the 3.14t leg.

Tests. tests/project/test_graalpy_shards_1963.py checks, through real pytest collection, that the shards cover every test of the tier exactly once. It fails against main (6 failed).

Part of #1963.

[JarryShaw]

GOOD TO GO at 7a910c9aa. A Sonnet review confirmed:

• Coverage. The pin collects the tier unsharded and through util/pytest_shards.py, then compares the sorted node ids, so a lost or doubled test fails it. Nested directories fall through to the directory shard or the remainder, and tests/conftest.py is unaffected.
• Exit status. timeout's 124/137 and the PIPESTATUS handling are unchanged, and pytest.log still receives every shard, so the crash and MemoryError counters see all of them. 3.14t runs pytest once, as before.
• Tests. The 1963, 1740, 1938 and 1732 pins pass.

One residual risk: a SIGINT landing between two shards is honoured only after the next shard starts, so that shard would run until the SIGKILL. The window is tiny.

[github-actions]

Coverage: 91.55% (unit tier, Python 3.15, 7a910c9aa, Unit Tests run success)

Package	Statements	Missed	Branches	Partial	Cover
pcapkit (top level)	104	4	20	2	95.16%
pcapkit/const	19597	881	2618	867	92.11%
pcapkit/corekit	2542	71	810	37	96.54%
pcapkit/dumpkit	276	3	100	3	98.40%
pcapkit/foundation	2930	23	1022	25	98.79%
pcapkit/interface	138	4	52	3	96.32%
pcapkit/protocols	17501	222	4612	187	98.13%
pcapkit/toolkit	872	36	298	10	95.21%
pcapkit/utilities	523	2	158	4	99.12%
pcapkit/vendor	4431	1995	1018	167	51.73%

Per-file detail: the coverage-html artifact of this run.