bug#30786: Save text properties in desktop

[Noam Postavsky <npostavs@gmail.com>]

Juri Linkov <juri@linkov.net> writes:

>>> Do you think this patch covers all possible unreadable cases on writing?
>>>
>>> +     ;; Get rid of unreadable text properties.
>>> +     (if (ignore-errors (read (format "%S" value)))
>>> +         (cons 'may value)
>>> +       (let ((copy (copy-sequence value)))
>>> +         (set-text-properties 0 (length copy) nil copy)
>>> +         (cons 'may copy))))
>>
>> I think it won't cover the case where an object's print syntax is only
>> readable by the current Emacs version, and not an earlier one.  To
>> handle that you'll need to call desktop--v2s recursively, like in the
>> vectorp and consp branches.
>
> I don't understand how calling desktop--v2s recursively will ensure its
> readability in earlier versions?  For example, if Emacs 27 will write to
> the desktop file ‘#("foo" 0 3 (bar 42))’ how Emacs 19 can read it?

Ah, I wasn't thinking of the #(...) specifically, rather about the
values of text properties, e.g.,

    (cl-defstruct foo f1 f2)
    (desktop--v2s (make-foo :f1 1 :f2 2)) ;=> (may . "Unprintable entity")

So desktop--v2s considers structs as unprintable, unlike read, hence
using desktop--v2s recursively instead of read will catch properties
that would otherwise be unreadable in Emacs 25.

If we want to support Emacs versions that don't even know about #(...)
then we can't write string properties at all, I guess.