[gordlin]

How the feature works? / How did you fix the issue?

Two changes that stop the Playwright suite reporting failures that are not about the
code under test. Every recent PR fails a large batch of specs, which makes the suite
useless as a gate — telling a real regression from the noise currently takes a manual
run-to-run diff.

The S-Group specs assume a default this fork changed

1a41ba32e ("clear history stack") changed the S-Group dialog's default type:

- const type = ssgroup.type || 'DAT';
+ const type = ssgroup.type || 'SRU';

Upstream's specs open the type dropdown by clicking the span that shows the current
type, matched on the literal text 'Data'. On this fork that span reads "SRU polymer",
so the click waits the full 60 seconds and the test fails. That accounts for 168 of
the click timeouts CI reports.

The helpers now use the s-group-type-input-span test id, which this repo already uses
in is-query-structure-selected.spec.ts and s-group-click-atom-bond.spec.ts. They no
longer depend on which type is the default, so they work on the fork and would work
upstream. The text parameter every caller passed as 'Data' is gone with it.

Measured locally on multiple-s-group-tool.spec.ts:

	before	after
click timeouts	11	0
run time	3.3 min	25 s

Shard 6 of 6 has never finished

Master's last three runs and every PR run hit the 120-minute cap on part (6, 6) and
were cancelled without printing a summary. A sixth of the suite therefore has no history
at all, so its failures cannot be compared against anything — that is why #115's CI
looked far worse than master's when the two are in fact almost identical.

Ten shards bring each one back under an hour.

Test plan

• multiple-s-group-tool.spec.ts locally, against a standalone build: 11 click timeouts
  to 0. The 11 remaining failures are A snapshot doesn't exist, from Playwright looking
  for -chromium-darwin.png on macOS; that cannot happen on Linux CI.
• npx tsc --noEmit and eslint clean over the changed specs.
• The first CI run on this branch is itself the check for the shard change.

Not included

About 150 screenshot comparison failures remain, concentrated in
toggle-explicit-hydrogens, image-files and calculate-cip-tool — the areas where the
fork deliberately changed what gets drawn (Indigo hydrogen label default, CIP
descriptors, S-group brackets) without regenerating upstream's committed snapshots.

Those need a scoped, reviewed regeneration, as a snapshot-only PR. Doing it before this
one lands would mean guessing at the list, since the current failure picture comes from
runs where a sixth of the suite never reported and every S-Group spec was timing out.

Check list

• unit-tests written
• e2e-tests written
• documentation updated
• PR name follows the pattern #1234 – issue name
• branch name doesn't contain '#'
• PR is linked with the issue
• base branch (master or release/xx) is correct
• task status changed to "Code review"
• reviewers are notified about the pull request

🤖 Generated with Claude Code