tmux/Q/Why does a session name containing a dot break every bare -t target?

• Why does a tmux session name containing a dot break every bare -t target?

  • AI Answer

    • Short answer: a -t target is parsed as session:window.pane, so a dot inside the session name is read as the window/pane separator. The lookup then runs against a window or a pane that does not exist, and the error names that window or pane rather than the session — which is why it reads as a missing session. Keep dots out of session names.

    • The parse, and where the name goes

      • session:window.pane is the target form. Every part is optional, so hay-.F is not rejected as malformed — it is accepted and split, with F taken as the pane.
      • The split happens on the first : and the first ., wherever they sit. A dot at the start of a suffix, in the middle of one, or in the middle of the name all break the same way; only the reported half changes.

    • Reproduced

      • Measured on 2026-08-24 Mon against tmux/v/3.7b, on an isolated socket rather than the server holding live work:
        • sh

          tmux -L probe new-session -d -s 'hay-.F'
          tmux -L probe list-sessions -F '#{session_name}'   # hay-.F — the session exists

      • target	exit	message
        has-session -t '=hay-.F'	1	can't find pane: F
        has-session -t 'hay-.F'	1	can't find pane: F
        list-windows -t '=hay-.F'	1	can't find pane: F
        new-window -d -t '=hay-.F'	1	can't specify pane here
        has-session -t '=hay-D.F'	1	can't find window: hay-D
        has-session -t '=hay.DF'	1	can't find window: hay
        has-session -t '=hay-DF'	0	—

      • The last row is the control: the same name with the dot removed answers cleanly.

    • It is not the = prefix, and not the shell

      • has-session -t 'hay-.F' fails identically to -t '=hay-.F', so exact-match plays no part. The target was quoted throughout, so zsh's equals expansion plays no part either — that is the separate fault on tmux/Q/Why does a tmux -t =session target fail in zsh?.
      • The two share a symptom and nothing else. Both report a name as not found while the session is sitting in list-sessions, and neither message contains the word session.

    • Why the error text is the expensive part

      • The message moves with the dot. hay-.F reports a missing pane; hay-D.F reports a missing window named hay-D, a string that appears nowhere. Two instances of one fault do not look alike, so recognising the second from having debugged the first does not happen.
      • new-window fails with can't specify pane here — an argument-shape complaint about a command that was given no pane, which points away from the name entirely.

    • has-session cannot answer, which is the real cost

      • A script's existence check returns 1, the same as for a session that is genuinely absent. Nothing downstream can tell a missing session from an unusable name, so the guard that would catch this is the thing the fault disables.

    • The workarounds, and why they are not a convention

      • A trailing colon disambiguates — -t '=hay-.F:' exits 0 — and an explicit window works too: -t '=hay-.F:0' lists panes normally.
      • Both are correct and neither is safe to build on. A naming scheme that holds only while everyone remembers a trailing colon fails at the first bare target anyone writes, including one written by a tool. An explicit window is also the case that hides the fault during testing: it is the one target shape a stray dot leaves alone, so a probe built from session:0 targets passes on every input it is given.

    • Sources

      • tmux(1) man page — COMMANDS — defines the session:window.pane target form the parse above follows.