At 2018-11-10T15:07:23-0200, Gabriel F. T. Gomes wrote: > Hi, > > On 10 Nov 2018, G. Branden Robinson wrote: > > >Package: bash-completion > >Version: 1:2.8-2 > >Followup-For: Bug #742466 > > > >I'm attaching a debdiff of a proposed NMU for this. I don't intend to > >actually NMU unless this bug threatens not to make it into the next > >Debian release (it's been around for several already :( ). > > It wouldn't be a problem, at all, if you submitted this as a NMU. > Anyhow, thanks for bringing this up for discussion first.
My pleasure. I see you recently took over the package, so I'm not
really inclined to preëmpt you, as exasperating as I have found the bug.
:)
> >I hope this will make it easy to apply and get shipping.
>
> It does, indeed. Thank you very much.
You're most welcome!
> >This bug has been complained about for years; see, e.g.:
>
> I'm sorry that this has been open for so long... I have been working
> on many ancient bugs recently, and I hope to fix many of them before
> the next Debian release. However, I started with those marked as
> important (I didn't question the reasons why a few bugs were marked as
> important, nor the correctness of the classification, because it
> predates my adoption of bash-completion).
Having maintained the XFree86 bug list long ago I appreciate what a task
this is. Starting with the high-severity bugs is the right thing to do.
And some bugs will have been miscategorized or insufficiently triaged;
it is the nature of the beast.
> Thank you very much for raising awereness for this bug... If you know
> of any other bugs that deserve immediate attention, please let me know
> and I'll glady focus on them.
For the most part, bash and its completions work as I expect, or
completions simply silently fail the way I got used to back in the days
before bash had programmable completion at all.
This was the only completion bug I can ever recall that upset me. :-O
> >Note that tab-completion within POSIX command substitution is still not
> >fully-fledged; with this fix, one gets command-name substitution after
> >'$(', but filename completion after entering a command does not take
> >place. One has to force it manually, say with M-/.
>
> I have tested this change and it does exactly what's described above,
> with one addition: after the command name substitution, nor filename,
> nor *subcommand* substitution take place. For instance:
>
> $ echo $(cv[TAB][TAB]
> cvlc cvs-debi cvs-switchroot
> cvs cvs-debrelease cvt
> cvs-debc cvs-debuild cvtsudoers
>
> as expected, but:
>
> $ echo $(cvs lo[TAB][TAB]
> (no output)
>
> when it should have completed `log' with `log '.
Yes, that's true. There is definitely still a bug, or a missing feature
here.
The instant issue is simply to get bash to stop spewing noise to stderr
during completion attempts.
> I agree and I'm preparing a new upload with this and other changes.
>
> May I attribute this change to you (as commit author) in the git
> repository for Debian's bash-completion. It would like like the
> attached patch.
I'm fine with that, but credit for identifying that stanza as requiring
deletion should go to Malte Skoruppa; he made the observation in June of
2015.
https://bugs.launchpad.net/ubuntu/+source/bash-completion/+bug/1312243/comments/8
Thank you for your efforts in getting bash-completion into good shape
for the buster release!
Regards,
Branden
signature.asc
Description: PGP signature
