RFC: Syntax for logging

RFC: Syntax for logging

[Daniel Jacobowitz]

Folks may remember the thread from a year ago:
  RFA: >, >>, and "tee" operators
  http://sources.redhat.com/ml/gdb-patches/2002-07/msg00458.html

I eventually decided that my prefered syntax was:
  redirect [-a] [FILE [COMMAND]]
  log [-a] [FILE [COMMAND]]
But people didn't care for the use of "-a".  I still like this syntax; it's
symmetric, and it allows clearly "transcript [-a]".  But it's pretty clear
to me that we won't reach a consensus on that.  I believe Fernando liked it
and Andrew didn't.

I believe the best alternative at this point is:
 set logging [redirect|log] [append|overwrite] FILE
 show logging
The defaults would be log,overwrite; they could be explicitly specified in
order to overwrite a log file named append, if one wanted to do that.

Comments, anyone?  Shall I repost the patch with that change?  I'd really
like to see this feature added.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer

RFC: Syntax for logging

[Doug Evans]

Daniel Jacobowitz writes:
 > Folks may remember the thread from a year ago:
 >   RFA: >, >>, and "tee" operators
 >   http://sources.redhat.com/ml/gdb-patches/2002-07/msg00458.html
 > 
 > I eventually decided that my prefered syntax was:
 >   redirect [-a] [FILE [COMMAND]]
 >   log [-a] [FILE [COMMAND]]
 > But people didn't care for the use of "-a".  I still like this syntax; it's
 > symmetric, and it allows clearly "transcript [-a]".  But it's pretty clear
 > to me that we won't reach a consensus on that.  I believe Fernando liked it
 > and Andrew didn't.
 > 
 > I believe the best alternative at this point is:
 >  set logging [redirect|log] [append|overwrite] FILE
 >  show logging
 > The defaults would be log,overwrite; they could be explicitly specified in
 > order to overwrite a log file named append, if one wanted to do that.
 > 
 > Comments, anyone?  Shall I repost the patch with that change?  I'd really
 > like to see this feature added.

There's one useful piece of functionality that isn't in "set logging".
Suppose I want to do a one-off command without disturbing the current setting?

This is an inherent problem in all set/show commands.
If one could have a version of show commands that outputted a value
that is an acceptable argument to the set command, and if one
has a facility to capture the output of commands and record them
in variables, then one would have a general solution, but that's
a bit of work (but not that much work ;-).

In pseudo-gdb code:

set $foo $`show -for-set logging`
set logging new-value
mumble
set logging $foo

fwiw, I don't think you should add a logging facility until
you know how you're going to solve the one-off request.
One don't have to solve it right away, but one should at least
have thought about it.
It needn't be solved by something so grandiose of course.
This is where redirect/log have an advantage though
I'm guessing one could come up with something simple that
allowed one-off's with "set logging".

Re: RFC: Syntax for logging

[Andrew Cagney]

> Folks may remember the thread from a year ago:
>   RFA: >, >>, and "tee" operators
>   http://sources.redhat.com/ml/gdb-patches/2002-07/msg00458.html
> 
> I eventually decided that my prefered syntax was:
>   redirect [-a] [FILE [COMMAND]]
>   log [-a] [FILE [COMMAND]]
> But people didn't care for the use of "-a".  I still like this syntax; it's
> symmetric, and it allows clearly "transcript [-a]".  But it's pretty clear
> to me that we won't reach a consensus on that.  I believe Fernando liked it
> and Andrew didn't.
> 
> I believe the best alternative at this point is:
>  set logging [redirect|log] [append|overwrite] FILE

>  show logging
> The defaults would be log,overwrite; they could be explicitly specified in
> order to overwrite a log file named append, if one wanted to do that.
> 
> Comments, anyone?  Shall I repost the patch with that change?  I'd really
> like to see this feature added.

Set show are consistent with the command line syntax.  How does one turn 
it off?

Andrew

Re: RFC: Syntax for logging

[Daniel Jacobowitz]

On Sun, Jun 22, 2003 at 02:04:29PM -0400, Andrew Cagney wrote:
> >Folks may remember the thread from a year ago:
> >  RFA: >, >>, and "tee" operators
> >  http://sources.redhat.com/ml/gdb-patches/2002-07/msg00458.html
> >
> >I eventually decided that my prefered syntax was:
> >  redirect [-a] [FILE [COMMAND]]
> >  log [-a] [FILE [COMMAND]]
> >But people didn't care for the use of "-a".  I still like this syntax; it's
> >symmetric, and it allows clearly "transcript [-a]".  But it's pretty clear
> >to me that we won't reach a consensus on that.  I believe Fernando liked it
> >and Andrew didn't.
> >
> >I believe the best alternative at this point is:
> > set logging [redirect|log] [append|overwrite] FILE
> 
> > show logging
> >The defaults would be log,overwrite; they could be explicitly specified in
> >order to overwrite a log file named append, if one wanted to do that.
> >
> >Comments, anyone?  Shall I repost the patch with that change?  I'd really
> >like to see this feature added.
> 
> Set show are consistent with the command line syntax.  How does one turn 
> it off?

Hmm, two options:
  set logging
  unset logging

I like "unset logging"; how about you?  The only thing we use unset for
at the moment is "unset environment" but I think it extends naturally.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer

Re: RFC: Syntax for logging

[Andrew Cagney]

> 
> Hmm, two options:
>   set logging

Not this.  It should behave like:

(gdb) set architecture
"set architecture" must be followed by "auto" or an architecture name.

>   unset logging
> 
> I like "unset logging"; how about you?  The only thing we use unset for
> at the moment is "unset environment" but I think it extends naturally.

so (slightly wierd) yes this.  Be sure to mention it in the help 
message.  Perhaphs, also accept ``set logging off'', and ``set logging 
on'' as those are the first things I'd try :-)

Hmm, just noticed, the options are optional, I was thinking that they 
were required.

	set logging [redirect|log] [append|overwrite] FILE

The syntax will lead to confusion with things like:

	set logging l
or
	set logging redirect a

suggest instead:

set logging on [FILE]
	default gdb.log, log, append?
set logging off
set logging file FILE
set logging append {true,false}
set logging redirect {true,false}

so that the [FILE] isn't context dependent.  Realisticly, I think people 
will only use:

	set logging on
	set logging off

Andrew

Re: RFC: Syntax for logging

[Daniel Jacobowitz]

On Sun, Jun 22, 2003 at 02:32:41PM -0400, Andrew Cagney wrote:
> >
> >Hmm, two options:
> >  set logging
> 
> Not this.  It should behave like:
> 
> (gdb) set architecture
> "set architecture" must be followed by "auto" or an architecture name.

OK.

> 
> >  unset logging
> >
> >I like "unset logging"; how about you?  The only thing we use unset for
> >at the moment is "unset environment" but I think it extends naturally.
> 
> so (slightly wierd) yes this.  Be sure to mention it in the help 
> message.  Perhaphs, also accept ``set logging off'', and ``set logging 
> on'' as those are the first things I'd try :-)
> 
> Hmm, just noticed, the options are optional, I was thinking that they 
> were required.
> 
> 	set logging [redirect|log] [append|overwrite] FILE
> 
> The syntax will lead to confusion with things like:
> 
> 	set logging l
> or
> 	set logging redirect a

Ugh, I'd forgotten about the abbreviations.

> suggest instead:
> 
> set logging on [FILE]
> 	default gdb.log, log, append?
> set logging off
> set logging file FILE
> set logging append {true,false}
> set logging redirect {true,false}
> 
> so that the [FILE] isn't context dependent.  Realisticly, I think people 
> will only use:
> 
> 	set logging on
> 	set logging off

That looks good.  I'll just ditch the unset idea.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer

Re: RFC: Syntax for logging

[Daniel Jacobowitz]

On Sun, Jun 22, 2003 at 02:41:41PM -0400, Daniel Jacobowitz wrote:
> On Sun, Jun 22, 2003 at 02:32:41PM -0400, Andrew Cagney wrote:
> > >
> > >Hmm, two options:
> > >  set logging
> > 
> > Not this.  It should behave like:
> > 
> > (gdb) set architecture
> > "set architecture" must be followed by "auto" or an architecture name.
> 
> OK.
> 
> > 
> > >  unset logging
> > >
> > >I like "unset logging"; how about you?  The only thing we use unset for
> > >at the moment is "unset environment" but I think it extends naturally.
> > 
> > so (slightly wierd) yes this.  Be sure to mention it in the help 
> > message.  Perhaphs, also accept ``set logging off'', and ``set logging 
> > on'' as those are the first things I'd try :-)
> > 
> > Hmm, just noticed, the options are optional, I was thinking that they 
> > were required.
> > 
> > 	set logging [redirect|log] [append|overwrite] FILE
> > 
> > The syntax will lead to confusion with things like:
> > 
> > 	set logging l
> > or
> > 	set logging redirect a
> 
> Ugh, I'd forgotten about the abbreviations.
> 
> > suggest instead:
> > 
> > set logging on [FILE]
> > 	default gdb.log, log, append?
> > set logging off
> > set logging file FILE
> > set logging append {true,false}
> > set logging redirect {true,false}
> > 
> > so that the [FILE] isn't context dependent.  Realisticly, I think people 
> > will only use:
> > 
> > 	set logging on
> > 	set logging off
> 
> That looks good.  I'll just ditch the unset idea.

Hmm, what do you think of this idea?

The above set commands, and also:
  log file CMD
"log" will obey "set logging append" and "set logging redirect".

An alternative, even simpler:
  log CMD
"log" will obey append, redirect, and file, but works even when logging
is off.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer