Control Sequences Support

[ErrorPower2001]

It seems psmux doesn't support Control Sequences. Are there any plans to do so?

The PSReadline module's prediction of history relies on ctlseq; currently, pwsh cannot predict history in psmux.

Version: 0.4.1

[marlocarlo]

Thanks for the report.

psmux does support and forward VT/ANSI control sequences -- it uses ConPTY which handles the full VT sequence pipeline. Things like colors, cursor movement, alternate screen, mouse reporting, and cursor shape changes all work.

However, PSReadLine's inline prediction feature is intentionally disabled inside psmux sessions. The reason: PSReadLine's prediction rendering writes VT sequences directly to the console in a way that races with ConPTY's output capture, causing display corruption (garbled text, misplaced cursor, flickering). This is a known conflict between PSReadLine's prediction implementation and terminal multiplexers -- tmux on WSL has similar issues with PSReadLine predictions.

Specifically, psmux runs this at shell startup:

Set-PSReadLineOption -PredictionSource None

What works: Tab completion, syntax highlighting, history search (Ctrl+R), all keybindings, $PSStyle.OutputRendering = 'Ansi' -- these all function normally.

What's disabled: PredictionSource History/HistoryAndPlugin (the inline gray suggestion text and F2 list view).

Could you clarify which specific control sequences or PSReadLine features you're seeing issues with? If it's specifically the inline prediction suggestions, that's the intentional disable described above. If there's something else not working, I'd like to investigate.

If you want to force predictions back on (with the caveat that you may see occasional rendering glitches), you can add this to your PowerShell profile when inside psmux:

if ($env:PSMUX_SESSION) {
    Set-PSReadLineOption -PredictionSource History -PredictionViewStyle InlineView
}

[ErrorPower2001]

Could you clarify which specific control sequences or PSReadLine features you're seeing issues with? If it's specifically the inline prediction suggestions, that's the intentional disable described above. If there's something else not working, I'd like to investigate.

Specifically, I have encountered the following two one(version 0.4.3) issues:

• Cursor Style Sequences: The control sequence CSI Ps SP q (following xterm documentation, where CSI is \e[, Ps is a single digit, and SP is a space) does not change the cursor style as expected.(fix in version 0.4.3)
• Character Attributes: The sequence CSI Pm m (where Pm represents multiple semicolon-separated positional parameters) is not functioning for several attributes. So far, I have tested and confirmed that the following do not work when outputting via Write-Host in pwsh:
  • Pm = 5: Blink
  • Pm = 7: Inverse
  • Pm = 8: Invisible/Hidden

These are the issues I have identified so far. I will continue to track the project and report any additional issues.

---

Test command:

Write-Host "Blink Text: `e[5mI am Blink`e[0m`nInverse Text: `e[7mI am Inverse`e[0m`nHidden Text: `e[8mI am Hidden`e[0m"

[marlocarlo]

Hey @ErrorPower2001, really appreciate the detailed follow-up with those specific SGR attributes. That kind of feedback is super helpful.

Good news: Blink, Inverse, and Hidden are all fixed now in commit 45ed4b4. Here's what was going on:

• Inverse (\e[7m): The rendering path was trying to manually swap fg/bg colors, but that does nothing when both are set to the terminal default. Switched it to use Modifier::REVERSED so the terminal emulator handles it properly.
• Blink (\e[5m) and Hidden (\e[8m): These modifier flags were just missing from the fast rendering path (rows_v2). Added SLOW_BLINK/RAPID_BLINK and HIDDEN flag handling there.

These went in with the cursor rendering improvements for Claude CLI. If you grab the latest master and rebuild, all three should work with your test command.

Thanks again for taking the time to test and report these. It genuinely helps make psmux better. And if you're finding it useful, I'd love it if you could spread the word or toss it a star. Helps more people discover the project!

[ErrorPower2001]

@marlocarlo Thank you for creating this wonderful project.

May I ask if I should close this issue, or keep it open so that I can track more ctlseq problems in the future?

[marlocarlo]

@marlocarlo Thank you for creating this wonderful project.

May I ask if I should close this issue, or keep it open so that I can track more ctlseq problems in the future?

I was waiting for your confirmation. We can close this for now. Feel free to open new issues. Always here.

And if Psmux helped you, please don't forget to share it with someone 🙏