[Bash-completion-devel] [Bug#764800]: Re: speed up bash completion init by > 30%...

[Linda A. Walsh]

Ville Skyttä wrote:
> No, because the time taken by have() within bash-completion is already
> 0, and we want to don't want to encourage anyone to use it any more by
> making it faster, we want it to go away. And while it's there, we do
> want to share code in have() and _have() (and the latter still does
> have a purpose).
> 
> But seriously, in case it was not clear already, I am personally not
> going to make the suggested changes. And I have a hunch that
> something's probably not right in your bash-completion 2.1 setup --
> have() shouldn't be that much of a performance issue unless you have
> lots of 3rd party completions that use it (which I respectfully doubt)
> and the function definition you posted isn't like that in 2.1...
----
Actually, due to recent file re-orgs, 'bash-completion' is on my
/usr/share/ now, which, as a shared resource, gets mounted separately
from root and /usr...with /etc/bash_completion.d being
dual-owned by bash and systemd.

But the culprit may be bash:
> grep completion /etc/bash.*
/etc/bash.bashrc:       if test -e /etc/bash_completion ; then
/etc/bash.bashrc:     . /etc/bash_completion
---
But /etc/bash_completion.sh which doesn't seem to be owned
by any package but sources /usr/share/bash-completion/bash_completion

Ah... so yes... I'd agree, I'm not sure where some of these
things are coming from... I'll have to spend some time on this--
of course where I really noticed the performance impact was on
cygwin -- which likely has a completely different set of
packages owning the same files....

So sorry for the premature bother... as soon as I saw
systemd's involvement, I realized something was off...