Re: auth-source.el and eieio

[David Engster <deng@randomsample.de>]

Reiner Steib writes:
> On Mon, Feb 14 2011, Ted Zlatanov wrote:
>
>> On Mon, 14 Feb 2011 14:20:53 -0800 Lars Ingebrigtsen <larsi@gnus.org> wrote: 
>>
>> LI> Ted Zlatanov <tzz@lifelogs.com> writes:
>>>> OK, so we can't drop 23.1 and older.  Are we bundling EIEIO with Gnus or
>>>> not?  If not, I'll fix auth-source.el to degrade semi-gracefully.
>>
>> LI> Would bundling EIEIO with Gnus have any other negative impact?  
>> LI> That is, are there incompatibilities in EIEIO itself that may
>> LI> freak out various XEmacs/Emacs instances that may (or may not)
>> LI> have other versions of EIEIO installed?
>>
>> LI> If this can be done cleanly (i.e., a lib directory that dgnushack.el
>> LI> figures out whether to add to the load path (or not)), then I think
>> LI> including it would be the way to go.
>
> It is likely to lead to load-path shadows with weird effects.
> [ Just imagine compiling Gnus from git on a system with several Emacs
> versions (e.g. Debian-based systems).  The user will probably compile
> with the oldest Emacs and run the compiled version of Gnus with all
> versions.  With the newer Emacs, Gnus' EIEIO files will probably
> shadow the newer ones from Emacs. ]

Installing an Emacs package behind the user's back is never a good
idea. Can't Gnus determine at run-time whether it has to put its own
EIEIO in the load-path or not?

>> That would work for me, assuming there are no version issues.  But I
>> don't know enough about EIEIO to say.  Reiner, WDYT?  Anyone else?
>
> I don't object, but it may lead to trouble.

I've PMed Eric and he doesn't see much of a problem. EIEIO hasn't seen
significant changes in roughly two years, so it's pretty stable. He's
fine with whatever we do with it.

-David