On Mon, 05 Mar 2001, Tom Roche wrote:
>> However, although CPerl seems happy to edit remote files without
>> problem, I can't seem to compile them.
[EMAIL PROTECTED] 06 Mar 2001 13:19:19 +0100
> That's because the function compile-internal uses start-process to
> start the compilation, and start-process is restricted to the
> current host.
<snip>
> Alas, I haven't been able to make much progress, here. I know what
> to do, but I haven't done it.
A feeling with which I am _very_ familiar :-)
> (And finding all the places where start-process or call-process
> should be replaced with process-file is a big job.)
Well, perhaps a quick hack for the interim ?-) Seriously, I'd like to
propose two options, being fully aware that I lack the elisp chops to
implement them--I'm trusting the competencies of die Trampvolk will
rectify my deficiencies.
* Shelling out. If one is running in-band (as, conveniently, I am :-)
one is already running a shell on the remote host. Could one then
redefine (e.g.) 'tramp-compile-in-band' to run a
remote-compile-command on the remote, redirect the (remote) output,
copy the output back to the local host, and display it in a
*compilation* buffer?
* Temp copying. Currently,
tramp.info
> In order to access remote files TRAMP needs to transfer their
> content to the local machine temporarily.
Since a file's content can be made local in the defined temp
directory, why not just have a 'tramp-compile' that makes a local
copy, and call 'compile' on that as normal? This would break
external references, of course, but for preliminary hack that would
suffice--it would be better than nothing. A more sophisticated hack
could perhaps, on encountering an external reference, prompt the
user to temp-copy the required file, or just go ahead and do it.
In comparison, shelling out would seem to be
+ cleaner
- requires an in-band method
+ would keep the file in its context (directory), which make
resolution of references to other files easier
? tougher to implement 'next-error'?
while temp coping would seem to be
- messier (copying of files leads to greater possibility of divergent
versioning)
+ easier to make work with the current 'compile'
? possibly harder to implement 'next-error'?
- harder to deal with external references
Am I missing anything?
Are either of these doable in reasonable time?
Or must we await process-file?
FWIW, [EMAIL PROTECTED]
