Sometimes the prompt isn't displayed when turning the panels off

[egmontkob]

Is there an existing issue for this?

• I have searched the existing issues

Midnight Commander version and build configuration

- 4.8.33
- 4.8.33-402-g723392f61

Operating system

Ubuntu 25.10

Is this issue reproducible using the latest version of Midnight Commander?

• I confirm the issue is still reproducible with the latest version of Midnight Commander

How to reproduce

Start mc, press Ctrl-O many times to toggle the panels on/off.

Expected behavior

When the panels are turned off, the subshell prompt should display.

Actual behavior

In about 10-20% of the cases, the subshell prompt isn't displayed.

(It sometimes appears briefly when you press another Ctrl-O to bring back the panels, before the panels actually appear. I guess this also depends on timings, and is subject to the terminal emulator actually painting that brief intermediate state.)

Additional context

The recent change of showing the first prompt in colors seems to be irrelevant. My pending big PR doesn't fix this either.

[zyv]

/cc @horkykuba

[horkykuba]

Yes, I know about this issue, and I even tried to find the cause recently. So far, I’ve only determined that the problem is somewhere near the beginning of the function src/subshell/common.c:invoke_subshell() or nearby.

[zyv]

Also interestingly, I think it's always reproducible for me on Alpine's console. I was thinking it has something to do with cons.saver, but apparently not, and it's the same issue that you are discussing.

[egmontkob]

If I force use_persistent_buffer = FALSE then the bug is gone. So it's somehow related to the persistent command line feature.

[egmontkob]

From another terminal, while mc's panels are shown, I peek into the terminal's normal screen using the command (use the outer terminal's number)

{ printf '\e[?47l'; sleep 1; printf '\e[?47h'; } > /dev/pts/<N>

This does not affect mc's behavior.

While peeking, I can already see if the next Ctrl-O will show the prompt or not. That is, things don't go wrong when hiding the panels, things already went wrong or right when the panels were shown.

---

Also possibly related:

/* Subshell: if set, then the prompt was not saved on CONSOLE_SAVE */
/* We need to paint it after CONSOLE_RESTORE, see: load_prompt */
gboolean update_subshell_prompt = FALSE;

Looks like the code should handle two different possible code paths, based on whether the normal screen already contains the prompt or not.

And maybe one of those code paths isn't handled properly.

---

Which is weird because during these Ctrl-O presses the prompt is not displayed again by the shell, so I have no idea when the execution can go in different ways. To be cont'd...

[egmontkob]

Related to #4809 perhaps?

[egmontkob]

When showing the panels, read_command_line_buffer() does a

// We send any remaining data to STDOUT before we finish.
flush_subshell (0, VISIBLY);

This flush_subshell() sometimes forwards \r\e[K\r + prompt + \r\e[K\r + prompt => when hiding the panels, the prompt shows up;

but sometimes forwards \r\e[K\r + prompt + \r\e[K\r => when hiding the panels, the prompt is missing.

mc is too fast / the shell is too slow. The prompt (printed twice; not sure why, perhaps an injected Ctrl-C is there in the game, maybe even two) hasn't fully arrived yet when mc gives up waiting for it.

(But why does it show up properly underneath the panels, where is that prompt taken from?)