On Thu, Apr 03, 2025 at 04:51:22PM +0200, Vincent Lefevre wrote:
I'm removing the "upstream" tag, because at

 https://github.com/kislyuk/argcomplete/issues/491#issuecomment-2774459991

the upstream author says:

 Argcomplete does not activate global completion by default and does
 not recommend that distributions activate it by default. Global
 completion is meant to be activated by the user. I am not a Debian
 maintainer, and I don't have any control over the choices that
 Debian package maintainers make when modifying the installation
 routine of this package.

So, in short, such completions must not be enabled by default for zsh
in Debian.

A bit more details:

Though the main issue introduced by the latest version 3.6.1 has now
been fixed upstream (in 3.6.2), the previous versions were already
badly interfering with user settings, and it was nearly impossible
for the user to easily find the cause. In my case, removing the
python3-argcomplete package (which had been installed automatically
via a "Recommends:" dependency) made issues I got in the past months
disappear; I hadn't reported them earlier because they were occurring
only on one of my machines and I wasn't able to find the cause.

There have been complaints about this package at least for Arch Linux,
where the global completion (i.e. changing the "-default-" completer)
was enabled by default like in Debian. Such completions have then been
removed in Arch Linux.

See also:
 https://www.zsh.org/mla/workers/2025/msg00129.html

While I don't have any specific disagreement with anything you say here, I'm reluctant to unilaterally revert what was apparently a deliberate change in response to a user request (see https://bugs.debian.org/944469) so close to the trixie freeze, particularly when I've really just been doing drive-by maintenance of this package.

Could you please file a separate bug report about this, and CC the reporter and fixer of #944469? We can then discuss whether e.g. the completions are appropriate for bash but not zsh, or should be made non-default again across the board, or something else.

To deal with the remainder of this bug, I'll upgrade to 3.6.2, which should at least be fairly non-controversial.

Thanks,

--
Colin Watson (he/him)                              [cjwat...@debian.org]

Reply via email to