On 12/05/17 17:14 PM, Stefan Monnier wrote:

> Looks good to me.
>
> [ While looking at this code, I notice that with cl-defmethod (which can
>   dispatch not only on EIEIO objects, but other object types as well)
>   we can now fold eieio-override-prin1 into object-write (i.e. make one
>   into an alias for the other).  ]

Yup, I think there's a fair amount of modularization that can be done. I
can make that part of the patch on #29541.

>> If this is okay, it's going into emacs-26, right?
>
> I think so, but we're pretty late in the pretest, so better ask Eli&John
> if they think it's OK for emacs-26.

Okay: John and Eli, this patch is to handle changes in object
implementation in Emacs 26: they are now implemented with records, which
can't be round-tripped with prin1 and read, they have to be written and
read specially. The eieio-persistent mechanism handles that, but
previously did not look inside hash tables or vectors to see if there
were objects in there.

That's causing failure in the pcache library, which other libraries
depend on. I'm hoping we can sneak this in under the wire so that pcache
doesn't fail in Emacs 26.

[...]

On 12/05/17 23:20 PM, Pierre Téchoueyres wrote:
> Eric Abrahamsen <eric@ericabrahamsen.net> writes:
>
>> On 12/05/17 14:02 PM, Stefan Monnier wrote:
>>>> Would it be too "heavy" to just copy the object and modify the copy?
>>>
>>> No, that's fine as well.
>>
>> Okay, the attached appears to work just fine. Pierre's recipe passes,
>> as do all the tests in eieio-test-persist. Pierre, maybe you could eval
>> this quickly and make sure that pcache works correctly again?
>
> Yes all my receipes tests are running fine. Many thanks.
> I've tested also with the unicode-fonts package
> (http://github.com/rolandwalker/unicode-fonts) from melpa.
> Everything seem to work fine.

Great! Thanks for testing.

Eric