Zsh Mailing List Archive
Messages sorted by:
Reverse Date,
Date,
Thread,
Author
RE: PATCH: cleanup in zle_tricky.c
- X-seq: zsh-workers 5917
- From: "Andrej Borsenkow" <borsenkow.msk@xxxxxx>
- To: "Sven Wischnowsky" <wischnow@xxxxxxxxxxxxxxxxxxxxxxx>, <zsh-workers@xxxxxxxxxxxxxx>
- Subject: RE: PATCH: cleanup in zle_tricky.c
- Date: Wed, 24 Mar 1999 13:41:23 +0300
- Importance: Normal
- In-reply-to: <199903230951.KAA18666@xxxxxxxxxxxxxxxxxxxxxxxxxxxx>
- Mailing-list: contact zsh-workers-help@xxxxxxxxxxxxxx; run by ezmlm
> >
> > Bsically, the completion does not work with ambiguous prefix.
> In my case I
> > had /u1/informix* as well - that way _path_files took another branch.
>
> Ah, I thought you had made sure that this didn't happen because we had
> this discussion already -- and hence I didn't test for it (the real
> problem was that without globcomplete there were no matches). At that
> time I had asked if we wanted the path to expand as far as possible in
> such cases.
> Since you perceived this as a bug, the patch below makes the answer
> `yes'.
>
> > In case of `.../*.suffix', the `/.*' was replaced by `/.', that 'course
> > could not match.
>
I am sorry for this misunderstanding - obviously, I was not clear enough
-( What I intended to say, was:
the bug with .../*.suffix (* removal) happened (or had effect) only in case
of ambiguous prefix (at least so my test results). I did not mean anything
else.
Now, that I have been working with this a bit more, the result of this patch
is somewhat unexpected. Now, the /a/b/c/d expands the prefix even if nothing
would match. Returning to my expamle - the bug with /u/i/s/*.h was, that
this does match, but the match failed (because of bug). And current
/u/i/s/*.c -> /usr/include/sys/*.c is a bit too much ...
> Ok, with the patch below and the ones from 5871, 5872, and 5875: does
> everything work as you want (or expect ;-) it now? Including those
> `test/{a1,b1,x1y}' problems you mentioned?
Yes, with your great completers all works as expected - with above exeption.
May I suggest, that this change would be backed off? Probably, with
configuration option (but I fail to see, when it would be useful). If direct
completion of /a/b/c/d fails, this is the job for approximate/correcting
completion then.
To be more precise - the change makes prefix to be expanded even if nothing
matches. I did not intend it, and did not ask for it.
cheers
/andrej
P.S. Sorry for turmoil this week. But when after every patch I get different
results after pressing TAB ... :-)
Messages sorted by:
Reverse Date,
Date,
Thread,
Author