[ESS-bugs] ess-mode 5.3.8; after-change-functions; RET at end of file

[Martin Maechler]

>>>>> "RoSp" == Rodney Sparapani <rsparapa at mcw.edu>
>>>>>     on Tue, 25 Nov 2008 09:12:42 -0600 writes:

    RoSp> Martin Maechler wrote:
    >>>>>>> "MT" == Markus Triska <markus.triska at gmx.at>
    >>>>>>> on Mon, 10 Nov 2008 21:58:30 +0100 writes:
    >>>>>>> 
    >> 
    MT> Hi Rodney,
    MT> Rodney Sparapani <rsparapa at mcw.edu> writes:
    >> 
    >> >> (if (fboundp 'make-local-hook) (progn
    >> >> (make-local-hook 'after-change-functions)
    >> >> (add-local-hook 'after-change-functions 'noweb-after-change-function))
    >> >> (add-hook 'after-change-functions 'noweb-after-change-function nil t))
    >> 
    MT> Yes, this is a good way to stay backwards compatible.
    >> 
    >> Well, this is the bug that I only now detected and which
    >> "completely" breaks Rnw-mode
    >> such that I had to retract the freshly released ESS 5.3.9 :
    >> 
    >> make-local-hook  is defined, but
    >> add-local-hook  does not exist  ===> error   
    >> 
    >> and the whole  Rnw-mode is  broken.
    >> 
    >> If I replace 'add-local-hook' by 'add-hook', things seem to
    >> work, but of course it's still the wrong if branch.
    >> The point would have been *NOT* to use make-local-hook
    >> unless in very outdated versions of emacs.
    >> 
    >> I am fixing it now (please don't touch it for the moment)...
    >> 
    >> Do we really still need to support Emacs 20.x ?
    >> [ The one I have here does no longer work -- it is not finding
    >> /usr/lib/X11/locale/locale.alias  and bails out ]
    >> 
    >> What about Xemacs? How old versions are supported there?
    >> I currently seem to have access to 21.4.17 on one platform and
    >> 21.4.21 on another.
    >> 
    >> Martin
    >> 
    RoSp> Hi Martin:

    RoSp> If make-local-hook is defined then add-local-hook must be defined as well!
    RoSp> On XEmacs, they both are found in subr.el I don't understand this. What
    RoSp> does GNU Emacs do in this case? Apparently, define one and not the other?!?
    RoSp> No wonder I didn't see that behavior here.

    RoSp> I have no opinion on which old versions need to be supported. That,
    RoSp> of course, can be easily changed by changing the documentation. However,
    RoSp> make-local-hook/add-local-hook appear to be the currently supported
    RoSp> method of handling this functionality on XEmacs (and inherited from
    RoSp> GNU Emacs AFAICT).

but have been outdated on emacs for quite some time,
with the functionality being part of (add-hook .) with a 4-th
argument 'local'

Emacs has make-local-hook with the warning about its
outdated-ness, but seems to never have add-local-hook.

    RoSp> FWIW, I don't recommend anything but the current
    RoSp> stable release of XEmacs.

Well if I start that (21.4.21) from one of our ubuntu servers,
I get a warning about character conversion, and then not even
the  Meta key works. 
If I use the older (x.y.17) from our departmental installation,
the Meta key works neither and I get three warnings about
character/font conversions, and "pink on white" (!!)
foreground-background combination, which is also pretty
off-putting.

Both is with using  'xemacs -q'.

    RoSp> However, I don't think you will have any
    RoSp> trouble with relatively recent versions like 21.4.17 and higher
    RoSp> (February 2005 and forward). Sun distributes 21.4.12 (January 2003)
    RoSp> with it's developer tools so it commonly seen in the wild, but it
    RoSp> suffers from the revert-buffer not updating Local Variables bug
    RoSp> <http://ess.r-project.org/Manual/ess.html#fn-1>. Other than that,
    RoSp> it works fine and many people here are using it to my chagrin.

I don't want to spend more time with Xemacs for the time being
and would be very grateful if you could check the current (svn!)
version of ESS with some versions of Xemacs.

Martin