Re: Tramp questions

[michael.albinus@gmx.de]

Gabor Greif <g.greif@lucent.com> writes:

> ## Reposting since my original question was posted via
> ## groups.google.com and did not make it to the newsgroup

[This posting didn't make it either; I've seen it on news.alcatel.com. I
suspect, that news server does not feed the article outside]

[Yes, you could find me in the A-LU X500 directory :-) ]

> Hi all,

Hi Gabor,

> first of all, big thank you to all who have developed and as
> maintaining emacs tramp. It is a truly great tool and have saved me
> from countless headaches!
>
> Some remarks and questions follow...
>
> 1) Is there a method which instead of sending back entire files, sends
> only the diffs to the last saved checkpoint of the file? Something
> like '/ssh+patch:...' ? I ask, because in my setup download is very
> fast, but upstream is unreliable for more than a few kB, so saving all
> but tiny files regularly hangs. Maybe the ssh tunnel to the outer
> world is defective or constrained, but I have no control over this
> aspect :-( A patch-based save would work for me, as I tend to save WIP
> files often, so the patches are very small.

Tramp itself does not support this mechanism. You could try the rsync or
rsyncc methods. According to the ChangeLog, there are some optimizations
since Tramp 2.1.17.

Emacs *could* support saving of file pieces, see function
`write-region'. But I believe, the arguments start and end are not used
as much in the codebase, and Tramp is just a library, which implements
such basic functions.

> 2) I have observed that 'M-x compile' will remotely run my (e.g.) make
> command. This is wonderful, and I became even more astonished seeing
> that clicking in compile errors in .c files even opens the remote file
> and positions the cursor. But sometimes the error parser gets confused
> and does not chop off the ":line:column" portion from the filename and
> tries to open e.g. 'foo.c:77:23', which - of course - does not exist.
> Is this a tramp-related problem? (I guess so, since locally all is
> good.)

I'm not aware of this problem. Maybe you could show an example.

> That's it, thanks for any hints in advance!
>
> Cheers,
>
>     Gabor

Best regards, Michael.