Stefan Monnier wrote:
>> This change does not affect the behavior of
>> `org-table-make-reference' because `eq' treats all instances
>> of the empty string as the same object anyway, e.g., 
>> `(eq (string-trim "aaabbb" "a+" "b+") "")' ==> t.
>
>Not quite so: (eq (string-to-multibyte "") "") => nil

Thank you for noticing this, Stefan.

I did many tests with `eq' and various kinds of generated empty 
strings before writing that commit message -- but, alas, I didn't 
think of testing with a multibyte string.  (I also looked in the 
documentation, and the fact that there was no explicit mention of 
all empty strings being the same object should have tipped me 
off!)

So, I believe this means there is an error in that commit message. 
I've attached the commitinfo below for easy reference, with 
'inline' disposition.

However, the code change itself remains correct... I think?  That 
is: the previous code presumably had a latent bug, in that there 
*could* have been times when the `eq' test would fail when 
comparing against a multibyte empty string.

I don't know much about how Org Mode gets merged into Emacs, nor 
whether there are opportunities for rebasing anywhere along the 
way in that process.  If there's a way to update that commit 
message, then I'd do so (I can post a new one here).  But if 
there's no such opportunity, then so be it: I'll be on the 
permanent record with a technical mistake in a log message.  I'm 
sure it wouldn't be the first time :-/.

Best regards,
-Karl