something really wrong with rpm -qf tab completion (24.0.95.1)

something really wrong with rpm -qf tab completion (24.0.95.1)

[Neal Becker]

emacs:  "24.0.95.1"
fedora linux 16 x86_64

In shell mode:
rpm -qf /usr/lib64/pyth
then hit tab.  cpu goes to 100%, emacs is unresponsive for a LONG time.
This problem seems to be new.

Seems to be triggered by rpm.  Replacing with e.g., 'ls' doesn't seem to produce 
the same symptoms.

 

Re: something really wrong with rpm -qf tab completion (24.0.95.1)

[Glenn Morris]

Neal Becker wrote:

> emacs:  "24.0.95.1"
> fedora linux 16 x86_64
>
> In shell mode:
> rpm -qf /usr/lib64/pyth
> then hit tab.  cpu goes to 100%, emacs is unresponsive for a LONG time.

Thanks for the report. This should really have gone to bug list though.

Anyway, it's doing `rpm -q -a', which can be slow, depending on your
system. It's dumb, in that:

it should probably give an explanatory message before running that command

it doesn't try to cache the result of this slow command in any way (cf
eg woman)

it should not actually be running this command when the "-f" argument is
used (compare with rpm -q -f TAB), it should be doing filename
completion

if you use `rpm -q -f', it does better, doing filename completion, but the
results it presents are all files, rather than just *.rpm files.

Re: something really wrong with rpm -qf tab completion (24.0.95.1)

[Glenn Morris]

Also,

rpm -qa --nodigest --nosignature

would be much faster (0.25s for me versus 2s for rpm -qa).
But then you have to jump through the hoop of testing that those options
are supported. Maybe they are old enough to be taken for granted.

Re: something really wrong with rpm -qf tab completion (24.0.95.1)

[Glenn Morris]

Glenn Morris wrote:

> if you use `rpm -q -f', it does better, doing filename completion, but the
> results it presents are all files, rather than just *.rpm files.

That was me being dumb, mixing up rpm -qf and rpm -qp.

Re: something really wrong with rpm -qf tab completion (24.0.95.1)

[Glenn Morris]

Maybe all that is needed for now is:

*** lisp/pcmpl-rpm.el	2012-01-19 07:21:25 +0000
--- lisp/pcmpl-rpm.el	2012-04-06 22:18:38 +0000
***************
*** 129,135 ****
  	      (pcomplete-opt "af.p(pcmpl-rpm-files)ilsdcvR")
  	    (if (pcomplete-test "-[^-]*p" 'first 1)
  		(pcomplete-here (pcmpl-rpm-files))
! 	      (pcomplete-here (pcmpl-rpm-packages))))))
         ((pcomplete-test "--pipe")
  	(pcomplete-here* (funcall pcomplete-command-completion-function)))
         ((pcomplete-test "--rmsource")
--- 129,137 ----
  	      (pcomplete-opt "af.p(pcmpl-rpm-files)ilsdcvR")
  	    (if (pcomplete-test "-[^-]*p" 'first 1)
  		(pcomplete-here (pcmpl-rpm-files))
!               (if (pcomplete-test "-[^-]*f" 'first 1)
!                   (pcomplete-here* (pcomplete-entries))
!                 (pcomplete-here (pcmpl-rpm-packages)))))))
         ((pcomplete-test "--pipe")
  	(pcomplete-here* (funcall pcomplete-command-completion-function)))
         ((pcomplete-test "--rmsource")

Re: something really wrong with rpm -qf tab completion (24.0.95.1)

[Glenn Morris]

Hopefully fixed in emacs-24 branch.

Re: something really wrong with rpm -qf tab completion (24.0.95.1)

[Neal Becker]

[-- Attachment #1: Type: text/plain, Size: 110 bytes --]

Thanks
On Apr 7, 2012 3:38 PM, "Glenn Morris" <rgm@gnu.org> wrote:

>
> Hopefully fixed in emacs-24 branch.
>

[-- Attachment #2: Type: text/html, Size: 344 bytes --]