Karl Kleinpaste <karl(a)justresearch.com> writes in xemacs-beta(a)xemacs.org:
 SL Baur <steve(a)xemacs.org> writes:
> It depends upon which version of the package-index file you are using.   
 Offhand, I didn't even know what package-index was in question. 
I
 already had the July 29 sumo tarballs installed, and was just toying
 with the package manager when I ran into this bug.  As it happens, I
 just now manually acquired the latest mailcrypt-2.02 package and
 installed it...and now when I run the exact same sequence in an
 "xemacs -q", it claims that entries are not signed at all.  This is
 notwithstanding the fact that I now have a package-index.LATEST.pgp
 with modtime of just moments ago, plus the prev. version as
 package-index.LATEST.pgp~, with modtime July 29 21:50. 
 ??? 
Right.  Jan Vroonhof has been having me do that for some number of
releases now to bypass mailcrypt signature checking problems.  I think 
the current situation is hopeless and I give up.
> You can either strip out the signature, or update to the most
recent
> package-index file on the FTP site (which includes a pointer to an
> updated version of the mailcrypt package). 
 As I said, I already got that later mailcrypt.  But I have no idea
how
 or why any of this has occurred.  Installing the latest mailcrypt
 shouldn't have affected the determination of *whether* the packages
 are PGP-signed.  In fact, now, after starting up the package manager
 (and telling it to continue even though entries are not signed), the
 mailcrypt library is not even loaded -- the variable mc-version is not
 known. 
 I am *so* confused. 
As am I.  I now have a pgp 2.6.3 binary built from sources off of
replay.com.  Should I try signing with that version instead?