Re: 5.0.3 +* -> git completion regression

[Bart Schaefer <schaefer@xxxxxxxxxxxxxxxx>]

On Dec 16,  5:37am, Phil Pennock wrote:
}
} % zsh -f
} ilmenite% emulate ksh
} ilmenite% case "$foo" in +*) echo snert;; esac
} zsh: bad pattern: +*
} %

Here's what I get when compiled with debugging turned on:

$ case foo in +*) echo bar;; esac
 BUG: character not handled in patcomppiece
zsh: bad pattern: +*
$ 

(Not sure why there's a leading space on the "BUG:" line, seems odd.)

Anyway it looks like we've triggered lookahead on "+" because of ksh
emulation, but then when the next character is not a paren it fails to
back up and reprocess the "+", or something to that effect.

This is all tangled up with "disable -p" changes, but selectively baking
out chunks that involve the kshchar flag hasn't so far shown anything.