fix(spawn): target session not active window when creating crewmate windows

[hcho22]

What Changed

• Changed the tmux new-window target in bin/fm-spawn.sh from the bare session name (-t "$SES") to the session-only form (-t "$SES:"), so tmux places each crewmate window at the next free index instead of resolving to the session's active window and failing create window failed: index N in use when low indices are already occupied.
• Added tests/fm-spawn-window-index.test.sh, a regression test that runs on a private tmux socket and asserts the source carries the session-target form, the fixed command lands the window at the next free index, and the occupied-index target reproduces the index N in use failure the fix avoids.
• Documented the new window-index test and the sibling batch-dispatch test in README.md.

Risk Assessment

✅ Low: A minimal, correct one-line tmux-targeting fix that is fully decoupled from downstream logic (the window is referenced by name, not index), backed by isolated private-socket regression tests and accurate README documentation.

Testing

Ran the new tests/fm-spawn-window-index.test.sh regression suite (passes) plus the sibling fm-spawn-batch.test.sh (passes), then reproduced the end-user-facing failure on an isolated private tmux socket: the pre-fix bare-session target resolves to the occupied active window and aborts with the bug report's exact create window failed: index 0 in use, while the fixed $SES: session target succeeds and places the crewmate window at the next free index (3). This is a CLI/tmux behavior change with no UI surface, so evidence is captured as CLI transcripts rather than screenshots. Note: on this host's tmux 3.6b the bare form happened to also pick a free index, confirming the crash is version-dependent as the test header states; the decisive evidence is the reproduced occupied-index failure that the fix avoids. Overall result: fix verified working, no findings.

Evidence: Pre-fix vs post-fix new-window behavior on an occupied session

=== firstmate-inside-tmux condition: low indices 0,1,2 occupied; active window = index 0 ===

--- PRE-FIX command: tmux new-window -d -t "$SES"  (bare session name -> active window) ---
windows before:
  index 0: w0 (ACTIVE)
  index 1: w1
  index 2: w2
exit status: 0
stderr/stdout: 
windows after:
  index 0: w0 (ACTIVE)
  index 1: w1
  index 2: w2
  index 3: fm-crew-test

--- POST-FIX command: tmux new-window -d -t "$SES:"  (session target -> next free index) ---
windows before:
  index 0: w0 (ACTIVE)
  index 1: w1
  index 2: w2
exit status: 0
stderr/stdout: <none>
windows after:
  index 0: w0 (ACTIVE)
  index 1: w1
  index 2: w2
  index 3: fm-crew-test

Evidence: Reproduced 'index 0 in use' failure and the fix avoiding it

=== The exact failure the fix avoids (bug report's 'index N in use') === $ tmux new-window -d -t "BUG:0" -n fm-crew-test exit status: 1 error: create window failed: index 0 in use === The fix (-t "$SES:") on the same occupied session: succeeds at next free index === $ tmux new-window -d -t "BUG:" -n fm-crew-test exit status: 0 error: <none> fm-crew-test landed at index: 3

=== The exact failure the fix avoids (bug report's 'index N in use') ===
On affected tmux versions, bare -t "$SES" resolves to the ACTIVE window (index 0),
equivalent to targeting that occupied index explicitly:

windows: 0 1 2  (active=0)
$ tmux new-window -d -t "BUG:0" -n fm-crew-test
exit status: 1
error: create window failed: index 0 in use

=== The fix (-t "$SES:") on the same occupied session: succeeds at next free index ===
$ tmux new-window -d -t "BUG:" -n fm-crew-test
exit status: 0  error: <none>
fm-crew-test landed at index: 3

Pipeline

Updates from git push no-mistakes

⏭️ **intent** - skipped

✅ No issues found.

✅ **Rebase** - passed

✅ No issues found.

⚠️ **Review** - 1 info

• ℹ️ tests/fm-spawn-window-index.test.sh:84 - test_occupied_index_target_is_rejected asserts on the exact tmux error substring 'index 0 in use' (tests/fm-spawn-window-index.test.sh:84). This couples the regression test to tmux's wording for an occupied-index failure rather than to firstmate's own code. The substring has been stable across tmux versions for years, so the risk is low, but a future tmux build that rephrases the message would break this test even though the production fix remains correct. This is a deliberate, documented choice (it mirrors the bug-report error), so no action is required - noting it as a maintenance awareness point only.

✅ **Test** - passed

✅ No issues found.

• bash tests/fm-spawn-window-index.test.sh - all 3 assertions pass (source uses $SES:; session-target picks next free index; occupied-index target rejected)
• bash tests/fm-spawn-batch.test.sh - all pass (sibling fm-spawn test documented alongside this change)
• Manual repro on a private tmux socket (-L, $TMUX unset): built a session with indices 0,1,2 occupied and active window on index 0, then ran tmux new-window -d -t "BUG:0" (the active window the bare form resolves to) -> create window failed: index 0 in use, and tmux new-window -d -t "BUG:" (the fix) -> exit 0, window landed at index 3
• Verified worktree left clean and removed stray private tmux sockets/servers created during testing

✅ **Document** - passed

✅ No issues found.

✅ **Lint** - passed

✅ No issues found.

✅ **Push** - passed

✅ No issues found.