shutdown-hook does not execute sync-mailbox (possible bug)

[rafaelrc7]

Hello, I'm trying to run multiple commands with a single hook. This sounds simple enough, but I was not able to do it nor find any information about online.

My ideia is to use shutdown-hook to first save changes with <sync-mailbox> and then run a shell script that synchronises the maildir itself. I am able to easily do each alone, writing either shutdown-hook 'exec sync-mailbox' or shutdown-hook 'echo `sync-script.sh`'. However, I was not able to run both, first sync-mailbox and then the script.

Is there a way that I could do that? Thanks.

[rafaelrc7]

Ok. I think I did manage to run both commands, by either doing:

shutdown-hook 'exec sync-mailbox `sync-script.sh`'

or

shutdown-hook 'exec sync-mailbox ; echo `sync-script.sh`'

However it still seems to not be working as expected... The sync-script (that is just a call to mbsync) detects no changes, so read emails are not updated, as if sync-mailbox had not been executed. However, if I run the hook two times, by closing, opening and closing it works, so the command is working but the mailbox is only being saved after the program exits.

How could I fix it?

[dorsiflexion]

In fact it seems that shutdown-hook 'exec sync-mailbox' is doing exactly nothing.

I'm trying to do something very similar. It doesn't even seem to be a race condition, because when doing

shutdown-hook 'exec sync-mailbox; echo `sleep 5`; echo `script.sh`'

it is still not updated when the script runs.

Btw. neomutt updates mailboxes when it quits. So if you are closing, opening, closing and it works, that is no indication that the hook actually did something.

I'd also greatly appreciate some advice on what's wrong here.

[rafaelrc7]

In fact it seems that shutdown-hook 'exec sync-mailbox' is doing exactly nothing.

I'm trying to do something very similar. It doesn't even seem to be a race condition, because when doing

shutdown-hook 'exec sync-mailbox; echo `sleep 5`; echo `script.sh`'

it is still not updated when the script runs.

Btw. neomutt updates mailboxes when it quits. So if you are closing, opening, closing and it works, that is no indication that the hook actually did something.

I'd also greatly appreciate some advice on what's wrong here.

Yes, I also tried using sleep in my hook to test if it was some race condition, and had the same results.

I ended up rebinding the exit key (q) to run the sync commands and then quit. It works. But it still seems to be something wrong with the shutdown-hook.

[roccoblues]

Hi,

to clarify: the shutdown-hook is executed right? The problem is that sync-mailbox doesn't work as expected in it?

My wild guess is that sync-mailbox works on the current mailbox and by the time the shutdown-hook is executed there's no "current mailbox" anymore. But we need @flatcap expertise here.

[dorsiflexion]

to clarify: the shutdown-hook is executed right? The problem is that sync-mailbox doesn't work as expected in it?

Exactly. If I do something like:

shutdown-hook 'exec sync-mailbox; echo `notify-send test`'

I'm getting the notification. It's just that sync-mailbox doesn't seem to have an effect.

[roccoblues]

I had a deeper look and the problem doesn't seem to be related to sync-mailbox. It looks like no exec command gets executed. With the following shutdown-hook

shutdown-hook "echo 'before'; exec sync-mailbox; echo 'after'"

and some additional logging I get:

[2024-03-18 08:44:11]<1> dlg_index() Got op OP_QUIT (198)
[2024-03-18 08:44:11]<5> mutt_startup_shutdown_hook() shutdown
[2024-03-18 08:44:11]<5> mutt_startup_shutdown_hook() hook: echo 'before'; exec sync-mailbox; echo 'after'
[2024-03-18 08:44:11]<5> parse_rc_buffer() buffer:
[2024-03-18 08:44:11]<1> parse_rc_buffer() NT_COMMAND: echo
[2024-03-18 08:44:11]<M> parse_echo() before
[2024-03-18 08:44:11]<5> msgwin_set_text() MW SET: 6, before
[2024-03-18 08:44:11]<5> msgwin_calc_rows() msgwin_calc_rows() => 1
[2024-03-18 08:44:11]<5> msgwin_recalc() recalc done, request WA_REPAINT
[2024-03-18 08:44:11]<5> msgwin_repaint() msgwin repaint done
[2024-03-18 08:44:11]<5> parse_rc_buffer() rc: 0
[2024-03-18 08:44:11]<N> notify_send() sending: NT_COMMAND/13
[2024-03-18 08:44:11]<N> send() send: 7, 0x12001f800
[2024-03-18 08:44:11]<1> parse_rc_buffer() NT_COMMAND: exec
[2024-03-18 08:44:11]<5> mutt_parse_exec() exec: echo 'before'; exec sync-mailbox; echo 'after'
[2024-03-18 08:44:11]<5> mutt_parse_exec() function: sync-mailbox
[2024-03-18 08:44:11]<5> mutt_push_macro_event() ch: 0 op: 174
[2024-03-18 08:44:11]<5> parse_rc_buffer() rc: 0
[2024-03-18 08:44:11]<N> notify_send() sending: NT_COMMAND/14
[2024-03-18 08:44:11]<N> send() send: 7, 0x12001f800
[2024-03-18 08:44:11]<1> parse_rc_buffer() NT_COMMAND: echo
[2024-03-18 08:44:11]<M> parse_echo() after
[2024-03-18 08:44:11]<5> msgwin_set_text() MW SET: 5, after

Notice how the exec sync-mailbox gets parsed correct but never executed. It's gets pushed as an "event" in mutt_push_macro_event() but that's it. I think that is because NeoMutt is no longer in the main loop and nothing is actually handling those events.

I'll see if I can find some changes related to that next.

[roccoblues]

Quick update here: we have a, not to ugly, PR that fixes the issue that exec functions aren't triggered in the hook.

However the behavior is confusing because the command queue is only triggered after the complete hook is parsed. Fixing this would be a way bigger change.

I suggest something like alias nm='neomutt && sync-script.sh' as a temporary workaround for the original issue.