bug#61962: 30.0.50; New trouble with symbols with positions

[Alan Mackenzie]

Hello, Mattias.

On Sun, Mar 12, 2023 at 18:30:51 +0100, Mattias Engdegård wrote:
> As promised earlier, I gave the safe-copy-tree code a good
> working-through. Testing revealed bugs but the new implementation
> shouldn't have them. It's internal for now as there seems to be no need
> for it elsewhere, which also permitted some gold-plating to be removed.
> The new code is also quite a bit faster.

I'm not at all happy with the changes you've made.  You've transformed a
general purpose utility into a special purpose restricted one.  It was me
that put the work in in the first place, and I wonder why.  It was a
substantial amount of work, and it would appear to have been for nothing.

Why did you not talk to me about the changes you were intending to make?
You've simply overridden my judgment with your own in cutting the scope
of the new function down.  Why?  I thoroughly disagree with you that no
general purpose copy-tree is needed (I've lamented its lack before now),
and I thoroughly disagree with you that vectors and records need never be
copied.

You say there were bugs with my version.  OK, thanks for correcting them,
but would you please identify exactly what the bugs were.  How else am I
supposed to learn?

You say your new version is "quite a bit" faster.  What does that mean?
A factor of 10?  A factor of 2?  20%?  10%?  How did you measure this
speed up, and what particular code change was responsible for it?
Thanks.

-- 
Alan Mackenzie (Nuremberg, Germany).