Re: ERT tests on Emacs22

Re: ERT tests on Emacs22

[Ted Zlatanov]

On Mon, 16 May 2011 15:16:15 +0200 David Engster <deng@randomsample.de> wrote: 

DE> trying to get the ERT tests to work on Emacs22, it occured to me that
DE> you're depending on the hashtable-print-readable feature, which is only
DE> available since Emacs 23.2? So I'm afraid you'll either have to restrict
DE> the registry to newer Emacsen, or you'll have to create an object-write
DE> method for registry-db which properly serializes hash-tables for older
DE> Emacsen...

Ugh.  Well, the gnus-registry is an optional package so we could simply
say it's not for older Emacsen and skip the ERT tests there.  The
package could, when loaded, complain if the feature is not available.

I'd really like to avoid writing a hashtable serializer for the sake of
Emacs 23.1 and older users.  It's been missing since the
gnus-registry.el rewrite and no one has complained, so I doubt it's a
big population of people who want the latest Gnus yet can't upgrade
their Emacs.

The next major release of Gnus should probably require 23.2 or newer,
anyhow.

Ted

Re: ERT tests on Emacs22

[David Engster]

Ted Zlatanov writes:
> On Mon, 16 May 2011 15:16:15 +0200 David Engster <deng@randomsample.de> wrote: 
> DE> trying to get the ERT tests to work on Emacs22, it occured to me that
> DE> you're depending on the hashtable-print-readable feature, which is only
> DE> available since Emacs 23.2? So I'm afraid you'll either have to restrict
> DE> the registry to newer Emacsen, or you'll have to create an object-write
> DE> method for registry-db which properly serializes hash-tables for older
> DE> Emacsen...
>
> Ugh.  Well, the gnus-registry is an optional package so we could
> simply say it's not for older Emacsen and skip the ERT tests there.
> The package could, when loaded, complain if the feature is not
> available.
>
> I'd really like to avoid writing a hashtable serializer for the sake
> of Emacs 23.1 and older users.  It's been missing since the
> gnus-registry.el rewrite and no one has complained, so I doubt it's a
> big population of people who want the latest Gnus yet can't upgrade
> their Emacs.

Perfectly fine with me; I'll drop the registry tests for Emacs22
then. BTW, XEmacs doesn't have the native serializer either; but as you
said, no one's complained yet...

-David

Re: ERT tests on Emacs22

[Ted Zlatanov]

On Wed, 18 May 2011 17:38:40 +0200 David Engster <deng@randomsample.de> wrote: 

DE> Perfectly fine with me; I'll drop the registry tests for Emacs22
DE> then. BTW, XEmacs doesn't have the native serializer either; but as you
DE> said, no one's complained yet...

I'm OK with dropping the registry tests for both those platforms.

I'm sure XEmacs has a native hashtable serializer, though.  EIEIO should
be patched to use it.

Ted

Re: ERT tests on Emacs22

[David Engster]

Ted Zlatanov writes:
> On Wed, 18 May 2011 17:38:40 +0200 David Engster <deng@randomsample.de> wrote: 
>
> DE> Perfectly fine with me; I'll drop the registry tests for Emacs22
> DE> then. BTW, XEmacs doesn't have the native serializer either; but as you
> DE> said, no one's complained yet...
>
> I'm OK with dropping the registry tests for both those platforms.
>
> I'm sure XEmacs has a native hashtable serializer, though.  EIEIO should
> be patched to use it.

EIEIO simply uses `prin1' to get a printed representation. If XEmacs has
a native hashtable serializer, it should work with this, but it seems it
doesn't:

#+begin_src lisp
(progn
  (setq db (make-hash-table)) 
  (puthash 'foo 5 db) 
  (with-temp-buffer
    (prin1 db (current-buffer))
    (goto-char (point-min))
    (read (current-buffer))))
#+end_src

--> Invalid read-syntax: Cannot read unreadable object

This is on XEmacs 21.5b29.

-David

Re: ERT tests on Emacs22

[Ted Zlatanov]

On Thu, 19 May 2011 08:51:48 +0200 David Engster <deng@randomsample.de> wrote: 

DE> EIEIO simply uses `prin1' to get a printed representation. If XEmacs has
DE> a native hashtable serializer, it should work with this, but it seems it
DE> doesn't:
DE> #+begin_src lisp
DE> (progn
DE>   (setq db (make-hash-table)) 
DE>   (puthash 'foo 5 db) 
DE>   (with-temp-buffer
DE>     (prin1 db (current-buffer))
DE>     (goto-char (point-min))
DE>     (read (current-buffer))))
DE> #+end_src

--> Invalid read-syntax: Cannot read unreadable object

DE> This is on XEmacs 21.5b29.

Stephen Turnbull said it does on emacs-devel a while ago when we
discussed the GNU Emacs hashtable serialization.  If an XEmacs user
could investigate this further, it would be helpful.

Ted

Gnus registry and Emacs <= 23.1 (was: ERT tests on Emacs22)

[Reiner Steib]

On Wed, May 18 2011, Ted Zlatanov wrote:

> I'd really like to avoid writing a hashtable serializer for the sake of
> Emacs 23.1 and older users.  It's been missing since the
> gnus-registry.el rewrite and no one has complained, so I doubt it's a
> big population of people who want the latest Gnus yet can't upgrade
> their Emacs.

You seem to assume that everybody using Gnus is using the registry.
That statement "nobody is using released Emacs with latest Gnus" is
not true.  Given the limited compatibility, I am glad that I didn't
change my setup (e.g. `nnmail-split-fancy-with-parent') to use the
registry years ago. :-(

Bye, Reiner.
-- 
       ,,,
      (o o)
---ooO-(_)-Ooo---  |  PGP key available  |  http://rsteib.home.pages.de/

Re: Gnus registry and Emacs <= 23.1

[Ted Zlatanov]

On Thu, 19 May 2011 08:25:42 +0200 Reiner Steib <reinersteib+gmane@imap.cc> wrote: 

RS> On Wed, May 18 2011, Ted Zlatanov wrote:
>> I'd really like to avoid writing a hashtable serializer for the sake of
>> Emacs 23.1 and older users.  It's been missing since the
>> gnus-registry.el rewrite and no one has complained, so I doubt it's a
>> big population of people who want the latest Gnus yet can't upgrade
>> their Emacs.

RS> You seem to assume that everybody using Gnus is using the registry.

You're right, I should have said "who want to use the registry and ...".

RS> That statement "nobody is using released Emacs with latest Gnus" is
RS> not true.

But I didn't say that. I said "no one has complained" (this is still
true: you are not complaining either since you don't use the registry)
and that I doubt there are many who can't upgrade.

RS> Given the limited compatibility, I am glad that I didn't change my
RS> setup (e.g. `nnmail-split-fancy-with-parent') to use the registry
RS> years ago. :-(

I don't want to write a hashtable serializer for the sake of users who
haven't upgraded their Emacs to 23.2 (released in 2008) and haven't even
asked for the feature.  I don't have time to spend on features no one
needs.  But if you're one of those users and you are asking me to do it
because you want to use that feature, that changes things.  Are you
asking?

Ted