Stephen Leake, 2007-04-18:

> It turns out that start-process arranges for both stderr and stdout to
> go to the buffer; I tested it. I'll suggest on the emacs-devel list
> that the manual be updated.

This is in the manual.  See (elisp) Output from Processes.


>>> Stephen Leake <[EMAIL PROTECTED]> writes:
>>>
>>>> So dvc-run-dvc-async explicitly spawns "sh", rather than the backend
>>>> executable.
> 
> So this is actually unnecessary.

Often, DVC needs to be able to distinguish the subprocess' stdout from 
its stderr so that it can parse the output received on stdout and 
display any errors received on stderr.  This becomes harder if you spawn 
the backend executable directly, since stdout and stderr will then be 
interleaved in the same buffer (or be sent to the same filter function). 
  However, there may be simple, backend-specific workarounds, such as 
looking for the "mtn: " prefixes that mtn seems to add to each line of 
its error messages.  This is not very reliable, though; I prefer the 
current approach.

For xmtn's interface to mtn automate stdio, the inability to distinguish 
stdout and stderr in Emacs' process filter functions is a problem with 
no elegant solution that I'm aware of.  Redirecting stderr to a file and 
reading it after the process has completed is acceptable for quick 
one-shot commands, but less easy for subprocesses that are potentially 
never terminated such as mtn automate stdio.

Christian.


_______________________________________________
Dvc-dev mailing list
[email protected]
https://mail.gna.org/listinfo/dvc-dev

Reply via email to