Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [dsdp-tm-dev] "Run a command" functionality

On Saturday, June 18, 2011 13:12:14 Vladimir Prus wrote:

It seems that there's general agreement this is a good idea,
but would it possible to get a definite confirmation from TM
maintainers that a patch to introduce such "Run a command"
functionality will be welcome, in principle?

We would not like to spend time making an upstream-suitable patch
if it is contrary to RSE philosophy in some way and will never
make it.

Thanks,

-- 
Vladimir Prus
CodeSourcery / Mentor Graphics
+7 (812) 677-68-40

> Hello,
> 
> in a few cases in our product, we're interested in just running a command
> remotely, getting its exit status and output. At the same time, it's
> explicitly not desirable to show the user any console window with the
> command output.
> 
> RSE provides shell service, which can be used to run a command, however
> this is good for interactive use, and has some problems when used to run a
> single command:
> 
> - The commands see they have a terminal, and starts to behave
> interactively, like asking the user some questions, which is bad in our
> case.
> - There's no clear way to obtain command error status. Per Anna's
> suggestion, we used
> 
> 	; echo "Exit status is $?"
> 
> but assumes such output is not shown by the program, and the last line of
> program output ends with newline, that we have a Linux target, and so on.
> - In addition to program output, there is output produced when logging in
> into the remote system, so we have to use some kludges to filter that out.
> 
> It seems to me that it would be best if this complexity (which seems
> specific to different connections methods) be incapsualted inside RSE. So,
> there should be an API to run a command, which can be implemented:
> 
> - Using SSH 'exec' channel for SSH connections
> - Using the above "echo" trick for telnet connections
> - Using whatever magic is necessary for DStore, or Windows, or other,
> connection types.
> 
> Does this seem to make sense? Shall I file a bug?
> 
> Thanks,



Back to the top