RE: Preview: portable dumper

[Drew Adams]

> > > Only rarely have we made changes that would prevent this.
> >
> > That used to be the case.  Now, nearly every major release is
> > incompatible, AFAICT.  As far as I can see, long gone are the
> > days when you could byte-compile in an older version and use
> > the byte-compiled code in a newer version.
> 
> I'm not aware of such problems.  AFAIK, we only made an incompatible
> change once, in Emacs 23.  The byte-compiled files compiled by later
               ^^^^^^^^^^^                           ^^^^^^^^^^^^^^^^^
> versions should all be compatible.  If you see a problem with that,
> please report it as a bug.

A change _in_ 23?  Or a change _after_ 23?  If the change
occurred in 23 then files compiled in 23 should be compatible
in later releases.  It should not be the case only that files
compiled in later releases are compatible with still later releases.

I cannot (always) byte-compile in 23 and use the result in
later releases, at least.  (It depends on the file.)

It may or may not represent a hard incompatibility.  It might
just be a change in a calling sequence or existence of some
macros.  I don't know, and I don't have the time to dig into it.

I do know that I do not hesitate to have conditional code that
(normally) is all that's needed, to deal with such changes,
and that that no longer suffices.

Are you suggesting that if someone can report a true
incompatibility introduced after Emacs 23 you will actually
change the byte compiler to remove it, and that you will
backport that fix to where it was introduced and later
releases?  (Otherwise, the fix doesn't really help.)

You can believe me or not.  I am used to conditionalizing
code to make it work with multiple releases.  And I have
always tried, for my own use, to byte-compile in the oldest
release possible, so that the result works for all of the
releases the particular library supports.  I have had to
stop compiling (except for testing, to pick up warnings)
some libraries because of what I've found to be
byte-compilation incompatibility.

Yes, my claim remains vague.  And yes, I'm anyway glad
to hear that there is some profession of still wanting
to keep compatibility for future releases.  That is a
large consolation, for which I am grateful.