Re: Bug triage

[Clint Adams <clint@xxxxxxx>]

On Mon, Dec 29, 2008 at 06:40:33PM -0500, Matt Wozniski wrote:
> Seems that not defining the variable when the function that defined it
> was interrupted would be best, if it's not too tough to implement.
> Regarding an empty variable as invalid is still better than allowing
> an empty variable, but depending on the code (which I haven't looked
> at) it might be able to create a variable holding only part of a valid
> cache.
> 
> Just something to think about.

I don't know.  How about

Index: Completion/Debian/Type/_deb_packages
===================================================================
RCS file: /cvsroot/zsh/zsh/Completion/Debian/Type/_deb_packages,v
retrieving revision 1.8
diff -u -r1.8 _deb_packages
--- Completion/Debian/Type/_deb_packages	2 Nov 2008 14:12:29 -0000	1.8
+++ Completion/Debian/Type/_deb_packages	30 Dec 2008 04:54:18 -0000
@@ -6,9 +6,11 @@
   if ( [[ ${+_deb_packages_cache_avail} -eq 0 ]] ||
       _cache_invalid DEBS_avail ) && ! _retrieve_cache DEBS_avail;
   then
-    _deb_packages_cache_avail=(
+    if _deb_packages_cache_avail=(
       ${(f)"$(apt-cache --generate pkgnames 2>/dev/null)"}
-    )
+    ); then
+      unset _deb_packages_cache_avail
+    fi
 
     _store_cache DEBS_avail _deb_packages_cache_avail
   fi