Re: BUG: Zsh loses history entries since 2015

[Bart Schaefer <schaefer@xxxxxxxxxxxxxxxx>]

On Thu, Apr 3, 2025 at 1:56 AM Michael Stapelberg
<stapelberg+zsh@xxxxxxxxxx> wrote:
>
> Yes. During my work day, I open many SSH connections from my laptop
> to my workstation. At the end of the work day, I kill the connection,
> which means all those zsh processes will be killed.

Given this, are you sure it's a SIGINT that's being received?

Typically in cases of a connection shutdown the signal would be
SIGHUP, potentially followed by a SIGKILL if it's a local terminal in
the window/session manager rather than a remote one via SSH.

I'm going to skip some things that Oliver addressed.

> > 2) Change behavior to append current shell history to the file as if
> > APPEND_HISTORY, without attempting to enforce SAVEHIST size or
> > duplicate expiration.  This is probably the option most likely to
>
> This approach would no longer rewrite history when exiting,
> so at which points would zsh rewrite history then?

To clarify, I meant that this behavior would be activated if and only
if the read/write was interrupted (or HUP'd, there's no recourse on
SIGKILL).

I did NOT mean to imply that this would be behavior of "normal" shell,
when NOT interrupted SAVEHIST would apply as it currently does.

> I have not checked what happens when savehistfile() is interrupted,
> but yes, any fix should probably cover both readhistfile() and savehistfile().

On closer look I am reminded that readhistfile() is called from within
savehistfile(), so this should be clarified as specifically "happens
during write-back".

However, I can't think of anything useful to do there ... with
hist-save-by-copy it should (already) just be leaving the ".new" file
behind and (because already dead) skipping the rename over the old
file.  So you probably haven't noticed if/when this happens after the
read is complete.

> > 3) Stop reading, but go on saving internal history per user
> > configuration.  Keeps all the current shell history but would lose
> > part of the existing history file.  Is that what happens now?
>
> I think so.

This is probably wrong (on INT) and not happening (on HUP).

> To be clear: the zsh processes that are killed in my scenario do not
> have any pending history to save

How do you know this?  Because if this could be programmatically
determined at time of exit, the entire history operation could be
optimized away.