[UX] pacman -Syu core update discards bash history

[Artoria2e5]

Description / Steps to reproduce the issue

1. Open a Window of MSYS2 MSYS
2. Do anything to populate bash history
3. Do pacman -Syu and make sure it's a core update
4. Do the update
5. Re-open the window

Expected behavior

I expect the history to be preserved, like with a normal close.

Actual behavior

The close is forced (for good reason) and no history is saved from the session.

Verification

• [X] I have verified that my MSYS2 is up-to-date before submitting the report (see https://www.msys2.org/docs/updating/)

(but you need to wait until it's outdated to reproduce it! oh or just make a fake update tarball.)

Windows Version

MSYS_NT-10.0-22624

Are you willing to submit a PR?

Yes, if we figure out the best way. A possible way is putting this in the default bashrc, but I am not sure it's "graceful" enough (what if it's not a core update? that would change behavior!). It also needs to cover other shells and so on.

pacman() {
  history -a;
  command pacman "$@";
}

Or maybe we make kill_all_other_msys_processes() do a normal sigterm in the MSYS2 context before the taskkill????

[lazka]

Yeah, that's not ideal right now, thanks

[lazka]

There are a few issues, open questions:

• bash ignores SIGTERM, so that doesn't work
• bash allows SIGHUP though (is that a good idea?)
• Does killing bash kill pacman running in it as well? Otherwise we don't get a chance of sending SIGKILL.

We could special case the parent processes of pacman somehow maybe?

[Artoria2e5]

Does killing bash kill pacman running in it as well

Ho, that's a good one! SIGHUP probably does that. I guess pacman can add a signal (SIGHUP, SIG_IGN); in the core upgrade bit?