Re: tramp (2.1.4); Corrupted files with syntax highliting

[Douglas K. Rand]

Michael> Could you, please, set `tramp-chunksize' to 500 manually, and
Michael> rerun the test? And maybe you even decrease the value step by
Michael> step. A comment in tramp.el says that the critical value is
Michael> 624 for FreeBSD, but there might be other circumstances with
Michael> you.

Doug> I've tried values of tramp-chunksize from 700 down to 10, and
Doug> all values caused corrupted files. (I didn't re-start Emacs
Doug> between changes, simply did (setq tramp-chunksize <value>) and
Doug> then modified and saved the file.

Michael> This wasn't enough. tramp-chunksize is evaluated during
Michael> connection setup, and set buffer local to *tramp/ssh
Michael> foo*. Either you change it being in that buffer, or you
Michael> delete the buffer after changing it globally somewhere
Michael> else. The connection is reestablished when you save the file
Michael> next time.

OK, I re-did the test, and it writes corrupted data less often with
smaller tramp-chunk sizes, but my experiments don't show 624 as a
threshold between working and non-working buffers. In fact, 10 was the
largest value I could get to work consistently.

For the test I did:

  (kill-buffer "*tramp/ssh address@hidden")
  (setq tramp-chunksize <value>)

I also ran the test described in the tramp-chunksize variable help
(after changing the step size to 10 and here are the results of a few
runs:

Bytes sent: 10  Bytes received: 0
Test does not work

Bytes sent: 10  Bytes received: 10
Bytes sent: 20  Bytes received: 20
Bytes sent: 30  Bytes received: 0
You should set `tramp-chunksize' to a maximum of 20

Bytes sent: 10  Bytes received: 10
Bytes sent: 20  Bytes received: 20
Bytes sent: 30  Bytes received: 30
Bytes sent: 40  Bytes received: 0
You should set `tramp-chunksize' to a maximum of 30

Bytes sent: 10  Bytes received: 0
Test does not work

Bytes sent: 10  Bytes received: 0
Test does not work

Bytes sent: 10  Bytes received: 10
Bytes sent: 20  Bytes received: 0
You should set `tramp-chunksize' to a maximum of 10

Clearly something is very screwed up for me.

Michael> PS: A workaround for you could be to use scp instead of
Michael> ssh. It doesn't use inline transfer, and doesn't enter the
Michael> process-send-string trap therefore.

Yup, scp works great!  Seems like an easier solution for now.

Let me know if you'd like me to try anything else.