Re: [cdt-debug-dev] Questions to be posted to GDB Mailing List regarding

[chanskw@xxxxxxxxxx]

Brian wrote:
I think it works marginally better the other way around, so
     <cli command>   produces CLI output, and
     -cli-command <cli command>   suppresses CLI output

I agree that it may be better for dealing with user defined commands and
sourced files.
However, do we break other front-ends that rely on entering plain gdb
command and
expecting to parse pure MI responses?
Or, are we going to force other front-ends to prefix all of their CLI
commands with -cli-command?

Output streaming may help in this case since all CLI-style should be
prefixed by "~" and other
frontends can just ignore those output if they cannot parse them.   I am
just not sure if other frontends
already have this capability to ignore the additional output or if we are
going to cause them some
grief by proposing this change.  Maybe we should also bring this to GDB
Mailing List for
discussion too.  We can suggest both alternatives in our question?

Proposal #1
===========
  <cli command> produces mixture of both CLI output and MI output -
continue to work as before
  -cli-command <cli command>  produces CLI output + MI responses - for
dealing with command line windows

In this case here, plain CLI command continues to produce the same output
as before.
It may or may not produce CLI output depending if it's a filler for missing
MI function.  With -cli-command,
gdb must produce CLI output for command line window to display.

*  but this is not a good solution to handle scripts in GDB
*  this does not require other frontends that use MI to change

Proposal #2
===========
  <cli command>  produces CLI output
  -cli-command <cli command>  suppresses CLI output

* this provides a better solution to handle scripts in GDB
* this may break frontends that currently use MI to interface with GDB.



Thanks,

Regards,
Samantha




|--------+------------------------------->
|        |          Brian                |
|        |          Thomson/Toronto/IBM@I|
|        |          BMCA                 |
|        |          Sent by:             |
|        |          cdt-debug-dev-admin@e|
|        |          clipse.org           |
|        |                               |
|        |                               |
|        |          04/22/2002 05:07 PM  |
|        |          Please respond to    |
|        |          cdt-debug-dev        |
|        |                               |
|--------+------------------------------->
  >-----------------------------------------------------------------------------------------------------------------------|
  |                                                                                                                       |
  |      To:     cdt-debug-dev@xxxxxxxxxxx                                                                                |
  |      cc:                                                                                                              |
  |      Subject:     Re: [cdt-debug-dev] Questions to be posted to GDB Mailing List regarding MI - comments?             |
  |                                                                                                                       |
  |                                                                                                                       |
  >-----------------------------------------------------------------------------------------------------------------------|



Samantha writes:
>    -cli-command <COMMAND>
> Entering this command tells the MI interpreter that this is
> really a CLI command and CLI output is expected from this command.

I think it works marginally better the other way around, so
     <cli command>   produces CLI output, and
     -cli-command <cli command>   suppresses CLI output
because it means you don't need to prefix what the user types with "
-cli-command".
I suspect it is better with sourced files and user-defined commands, none
of which
will have -cli-command in them.

So, to recap, the objectives are
     let users interact with CLI
     let GUI use MI
     let GUI also use CLI where MI is incomplete, but in this case we don't
want CLI output
     add MI events so GUI can keep current, even after CLI commands from
console or scripts

where we currently accomplish the last objective by re-querying the state
of everything after
any console activity.


Brian Thomson

_______________________________________________
cdt-debug-dev mailing list
cdt-debug-dev@xxxxxxxxxxx
http://dev.eclipse.org/mailman/listinfo/cdt-debug-dev