GEMMA DEVELOPER AGENTSAVED EVIDENCE
← Dataset Exploration
ARCHIVED TRAINING / tophat/syrupy

tophat__syrupy-761

This is a preserved training example. Membership in the current training wave has not been verified.
tophat/syrupy · Task 159 of 169

Problem statement

Syrupy is recognizing snapshot as not being used (4.0.1)
**Describe the bug**

This is probably a regression in the latest release. I didn't observe the behavior in 4.0.0:

1. On a new test with no snapshot (using the `json` extension), I get the expected behaviour:
> 1 snapshot failed
> Snapshot 'TestFoo.test_foo' does not exist!
2. With the first `pytest --snapshot-update` (this is where the first signs of the bug is):
> 1 snapshot generated. 1 unused snapshot deleted.
Deleted TestFoo (tests/foo/__snapshots__/test_foo/TestFoo.test_foo.json)

Why does it output `Deleted ....`, that isn't the expected behaviour (the deletion doesn't happen though)
3. Another follow up when running `pytest`:
> 1 snapshot passed. 1 snapshot unused.
Re-run pytest with --snapshot-update to delete unused snapshots.

Again, that isn't correct. It gets worse though:
4. `pytest --snapshot-update`:
This time the deletion *will* happen hence the next run of `pytest` will fail since it won't find a snapshot.

# Env: 

Syrupy 4.0.1
Python 3.10


Going back to 4.0.0 solved the issue.

Training membership

opd_e4b_pure_v6a opd_e4b_pure_v6b

Task issues extracted from preserved verified_rft_messages training examples. This is not a verified current RL wave manifest.

Task and snapshot provenance

Snapshot captured 2026-11-12T17:06:59.476467+00:00 · Autoresearch-Arena/gemma-4-developer-agent

{
  "datasets": [
    "opd_e4b_pure_v6a",
    "opd_e4b_pure_v6b"
  ],
  "examples": 4,
  "instance_id": "tophat__syrupy-761",
  "problem_statement": "Syrupy is recognizing snapshot as not being used (4.0.1)\n**Describe the bug**\r\n\r\nThis is probably a regression in the latest release. I didn't observe the behavior in 4.0.0:\r\n\r\n1. On a new test with no snapshot (using the `json` extension), I get the expected behaviour:\r\n> 1 snapshot failed\r\n> Snapshot 'TestFoo.test_foo' does not exist!\r\n2. With the first `pytest --snapshot-update` (this is where the first signs of the bug is):\r\n> 1 snapshot generated. 1 unused snapshot deleted.\r\nDeleted TestFoo (tests/foo/__snapshots__/test_foo/TestFoo.test_foo.json)\r\n\r\nWhy does it output `Deleted ....`, that isn't the expected behaviour (the deletion doesn't happen though)\r\n3. Another follow up when running `pytest`:\r\n> 1 snapshot passed. 1 snapshot unused.\r\nRe-run pytest with --snapshot-update to delete unused snapshots.\r\n\r\nAgain, that isn't correct. It gets worse though:\r\n4. `pytest --snapshot-update`:\r\nThis time the deletion *will* happen hence the next run of `pytest` will fail since it won't find a snapshot.\r\n\r\n# Env: \r\n\r\nSyrupy 4.0.1\r\nPython 3.10\r\n\r\n\r\nGoing back to 4.0.0 solved the issue.",
  "repo": "tophat/syrupy",
  "source": "v2/experiments/train_data/opd_e4b_pure_v6a/verified_rft_messages.jsonl",
  "source_sha256": "f045b7ca5b3883a1e2eaee81e1c02a4a1b53dd7b7b36e98bacb2ba5823a15148"
}
{
  "captured_at": "2026-11-12T17:06:59.476467+00:00",
  "membership": "archived training",
  "membership_evidence": "Task issues extracted from preserved verified_rft_messages training examples. This is not a verified current RL wave manifest.",
  "name": "Preserved post-training task archive",
  "source": "Autoresearch-Arena/gemma-4-developer-agent",
  "source_revision": "016496de222aaafca0d4ee4c18efc63ba616f76e",
  "sources": [
    {
      "path": "v2/experiments/train_data/opd_e4b_cispo_v5/verified_rft_messages.jsonl",
      "sha256": "a43610ad0dfb6ace95375cfae09000f41afc0338cc925af870a4ade7a9b20c4c"
    },
    {
      "path": "v2/experiments/train_data/opd_e4b_hybrid_v3/verified_rft_messages.jsonl",
      "sha256": "bd326303eb127e8e5e08f03daf97bb88d79d19709d59f3f93f348bc56686a8df"
    },
    {
      "path": "v2/experiments/train_data/opd_e4b_onpolicy_v3/verified_rft_messages.jsonl",
      "sha256": "960c680e797e823a0d3a22f4b1d5ef60a72971373d703b877340b0c9126aaf68"
    },
    {
      "path": "v2/experiments/train_data/opd_e4b_pure_v6a/verified_rft_messages.jsonl",
      "sha256": "f045b7ca5b3883a1e2eaee81e1c02a4a1b53dd7b7b36e98bacb2ba5823a15148"
    },
    {
      "path": "v2/experiments/train_data/opd_e4b_pure_v6b/verified_rft_messages.jsonl",
      "sha256": "0f62460a9706eea5e1605255896f191ab00cb5706c5f1f2e1c293e9a718218dc"
    },
    {
      "path": "v2/experiments/train_data/opd_e4b_pure_v7/verified_rft_messages.jsonl",
      "sha256": "e37f7bd675322f3f81500b73559d60d5f602ca0537bf19f5cb06332c728c7b72"
    },
    {
      "path": "v2/experiments/train_data/opd_e4b_pure_v7b/verified_rft_messages.jsonl",
      "sha256": "b864bdc8695183433eef6e31cb8d33fad44ba3aded2170c1ade4ad5b2944ce23"
    },
    {
      "path": "v2/experiments/train_data/opd_e4b_self_audit_v4/verified_rft_messages.jsonl",
      "sha256": "0026023745afb8f3dc1c176dc7f7171157b4ad5a66656b524af218e8e19ceee6"
    },
    {
      "path": "v2/experiments/train_data/rft_e4b_aligned_v2/verified_rft_messages.jsonl",
      "sha256": "92a4515fee9d152b0d4c510ee3ce4caaaffe2003ba7584469f8027f174aab18c"
    },
    {
      "path": "v2/experiments/train_data/rft_v1/verified_rft_messages.jsonl",
      "sha256": "a42453ccd7ac7680b0a03865e8e7080bc8bf74c8d9d193b440c441ed46f8512b"
    },
    {
      "path": "v2/experiments/train_data/rft_v2/verified_rft_messages.jsonl",
      "sha256": "18f4b3bd7e651f83c6acab479c9475e3fa50dc652a384f309e97146983ac85ba"
    }
  ]
}