Re: Garbage in files written with tramp

[Kai Großjohann]

Vincent Bernat <address@hidden> writes:

> OoO La nuit  ayant déjà recouvert d'encre ce jour  du lundi 07 février
> 2005, vers 23:38, Michael Albinus <address@hidden> disait:
>
>>> - Emacs   21.3.50  with   multi-tty   patch  (the   last  change   in
>>> lisp/Changelog is dated 2005-02-04 by Jay Belanger)
>>> - Tramp  from this  CVS  (but I  have  also tried  Tramp 2.0.47  from
>>> Debian)
>>> - metamail 2.7-46
>>> 
>>> Note that  I have  successfully decoded the  data of the  debug buffer
>>> with mimencode -u -b too.
>
>> Hmm, everything looks fine. Maybe you could check which version of
>> mimencode is present to your root account, because it seems to fail
>> decoding there.
>
> I have modified mimencode so that it dumps what it reads from stdin in
> a file. I have then diffed the file and the output in debug buffer. It
> seems  that small  chunks are  simply  discarded. The  error seems  to
> happen every 50 lines of base64 data.

Sorry for jumping in so late.  Are you reading or writing a file, and
is the mimencode running locally or remotely?  (Well, you are using
sudo, I think, so it's difficult to say which one is local and which
one is remote.  But I hope you understand what I mean.)

Frobbing the variable tramp-chunksize means that Emacs sends its
output to the shell more slowly.  Hm.

Also, the value 2000 works for you, but normally, people need to
choose a value like 500.  So perhaps the test doesn't catch this
situation: for you, it works to send 2000 bytes at a time, but the
test only tries to send 1000.

Perhaps you could try the test with a larger number of bytes, say
10,000?

Kai