[RFA] "Help" command and completion

[RFA] "Help" command and completion

[Michael Snyder]

Hmmm... currently the help command uses the default completer, 
which completes on symbols.  That's definitely not right.
Ideally, we want to complete on commands or something, but
substituting the noop complete would definitely be better than
completing on symbols.

OK to check in?

2002-01-04  Michael Snyder  <msnyder@redhat.com>

	* cli/cli-cmds.c (init_cli_cmds): The "help" command should not
	use the default completer.  Substitute noop_completer, although
	a command completer would be better.

Index: cli/cli-cmds.c
===================================================================
RCS file: /cvs/src/src/gdb/cli/cli-cmds.c,v
retrieving revision 1.10
diff -c -3 -p -r1.10 cli-cmds.c
*** cli-cmds.c	2001/09/01 21:38:05	1.10
--- cli-cmds.c	2002/01/05 02:47:58
*************** when gdb is started.", &cmdlist);
*** 668,674 ****
    c->completer = filename_completer;
  
    add_com ("quit", class_support, quit_command, "Exit gdb.");
!   add_com ("help", class_support, help_command, "Print list of commands.");
    add_com_alias ("q", "quit", class_support, 1);
    add_com_alias ("h", "help", class_support, 1);
  
--- 668,676 ----
    c->completer = filename_completer;
  
    add_com ("quit", class_support, quit_command, "Exit gdb.");
!   c = add_com ("help", class_support, help_command, 
! 	       "Print list of commands.");
!   c->completer = noop_completer;
    add_com_alias ("q", "quit", class_support, 1);
    add_com_alias ("h", "help", class_support, 1);
  

Re: [RFA] "Help" command and completion

[Eli Zaretskii]

> Date: Fri, 4 Jan 2002 18:47:17 -0800
> From: Michael Snyder <msnyder@cygnus.com>
> 
> Hmmm... currently the help command uses the default completer, 
> which completes on symbols.  That's definitely not right.
> Ideally, we want to complete on commands or something, but
> substituting the noop complete would definitely be better than
> completing on symbols.

How is no completion better than completion on symbols?

Re: [RFA] "Help" command and completion

[Michael Snyder]

Eli Zaretskii wrote:
> 
> > Date: Fri, 4 Jan 2002 18:47:17 -0800
> > From: Michael Snyder <msnyder@cygnus.com>
> >
> > Hmmm... currently the help command uses the default completer,
> > which completes on symbols.  That's definitely not right.
> > Ideally, we want to complete on commands or something, but
> > substituting the noop complete would definitely be better than
> > completing on symbols.
> 
> How is no completion better than completion on symbols?

Because the symbols in the program being debugged
can't possibly have anything to do with gdb commands, 
and therefore can't possibly be valid arguments for "help".
Not even if you're debugging gdb.

Re: [RFA] "Help" command and completion

[Michael Snyder]

Michael Snyder wrote:
> 
> Eli Zaretskii wrote:
> >
> > > Date: Fri, 4 Jan 2002 18:47:17 -0800
> > > From: Michael Snyder <msnyder@cygnus.com>
> > >
> > > Hmmm... currently the help command uses the default completer,
> > > which completes on symbols.  That's definitely not right.
> > > Ideally, we want to complete on commands or something, but
> > > substituting the noop complete would definitely be better than
> > > completing on symbols.
> >
> > How is no completion better than completion on symbols?
> 
> Because the symbols in the program being debugged
> can't possibly have anything to do with gdb commands,
> and therefore can't possibly be valid arguments for "help".
> Not even if you're debugging gdb.

OK to commit?

Re: [RFA] "Help" command and completion

[Eli Zaretskii]

On Tue, 8 Jan 2002, Michael Snyder wrote:

> Michael Snyder wrote:
> > 
> > Eli Zaretskii wrote:
> > >
> > > How is no completion better than completion on symbols?
> > 
> > Because the symbols in the program being debugged
> > can't possibly have anything to do with gdb commands,
> > and therefore can't possibly be valid arguments for "help".
> > Not even if you're debugging gdb.
> 
> OK to commit?

Personally, I'd rather have it complete on commands, like you said it
should.  We already have the function to do that: complete_on_cmdlist,
so it shouldn't be too hard to make that happen.

In fact, I just took a quick attempt on that; the patch is below.  It 
seems to work for me.  (The diffs are against stock GDB 5.1, since
that's what I happen to have on the machine I have here.)

In any case, IIRC completion is Fernando's land, so I guess we need to
wait for his word.


2002-01-09  Eli Zaretskii  <eliz@is.elta.co.il>

	* completer.c (command_completer): New function.

	* completer.h <command_completer>: Add prototype.

	* cli/cli-cmds.c (init_cli_cmds): Make command_completer be the
	completer for the "help" command.

--- gdb/completer.c~0	Sun Jul 15 21:17:46 2001
+++ gdb/completer.c	Wed Jan  9 09:05:48 2002
@@ -339,6 +339,14 @@ location_completer (char *text, char *wo
   return list;
 }
 
+/* Complete on command names.  Used by "help".  */
+char **
+command_completer (char *text, char *word)
+{
+  return complete_on_cmdlist (cmdlist, text, word);
+}
+
+
 /* Here are some useful test cases for completion.  FIXME: These should
    be put in the test suite.  They should be tested with both M-? and TAB.
 
--- gdb/completer.h~0	Sun Jul 15 21:17:46 2001
+++ gdb/completer.h	Wed Jan  9 08:52:50 2002
@@ -29,6 +29,8 @@
 
 extern char **location_completer (char *, char *);
 
+extern char **command_completer (char *, char *);
+
 extern char *get_gdb_completer_word_break_characters (void); 
 
 extern char *get_gdb_completer_quote_characters (void);
--- gdb/cli/cli-cmds.c~0	Mon Jul 16 16:33:46 2001
+++ gdb/cli/cli-cmds.c	Wed Jan  9 08:58:40 2002
@@ -668,7 +668,8 @@ when gdb is started.", &cmdlist);
   c->completer = filename_completer;
 
   add_com ("quit", class_support, quit_command, "Exit gdb.");
-  add_com ("help", class_support, help_command, "Print list of commands.");
+  c = add_com ("help", class_support, help_command, "Print list of commands.");
+  c->completer = command_completer;
   add_com_alias ("q", "quit", class_support, 1);
   add_com_alias ("h", "help", class_support, 1);
 

Re: [RFA] "Help" command and completion

[Fernando Nasser]

Sorry folks, I just came back from vacations and I still have to catch
up with all the e-mail.   I saw this one by chance while searching for
another thing...

Yes, I agree 100% with you both. 

I suggest that to proper document how we got here, Michael should check
his patch first and then Eli checks his in. 

Regards to all and Happy New Year.

Fernando


Eli Zaretskii wrote:
> 
> On Tue, 8 Jan 2002, Michael Snyder wrote:
> 
> > Michael Snyder wrote:
> > >
> > > Eli Zaretskii wrote:
> > > >
> > > > How is no completion better than completion on symbols?
> > >
> > > Because the symbols in the program being debugged
> > > can't possibly have anything to do with gdb commands,
> > > and therefore can't possibly be valid arguments for "help".
> > > Not even if you're debugging gdb.
> >
> > OK to commit?
> 
> Personally, I'd rather have it complete on commands, like you said it
> should.  We already have the function to do that: complete_on_cmdlist,
> so it shouldn't be too hard to make that happen.
> 
> In fact, I just took a quick attempt on that; the patch is below.  It
> seems to work for me.  (The diffs are against stock GDB 5.1, since
> that's what I happen to have on the machine I have here.)
> 
> In any case, IIRC completion is Fernando's land, so I guess we need to
> wait for his word.
> 
> 2002-01-09  Eli Zaretskii  <eliz@is.elta.co.il>
> 
>         * completer.c (command_completer): New function.
> 
>         * completer.h <command_completer>: Add prototype.
> 
>         * cli/cli-cmds.c (init_cli_cmds): Make command_completer be the
>         completer for the "help" command.
> 
> --- gdb/completer.c~0   Sun Jul 15 21:17:46 2001
> +++ gdb/completer.c     Wed Jan  9 09:05:48 2002
> @@ -339,6 +339,14 @@ location_completer (char *text, char *wo
>    return list;
>  }
> 
> +/* Complete on command names.  Used by "help".  */
> +char **
> +command_completer (char *text, char *word)
> +{
> +  return complete_on_cmdlist (cmdlist, text, word);
> +}
> +
> +
>  /* Here are some useful test cases for completion.  FIXME: These should
>     be put in the test suite.  They should be tested with both M-? and TAB.
> 
> --- gdb/completer.h~0   Sun Jul 15 21:17:46 2001
> +++ gdb/completer.h     Wed Jan  9 08:52:50 2002
> @@ -29,6 +29,8 @@
> 
>  extern char **location_completer (char *, char *);
> 
> +extern char **command_completer (char *, char *);
> +
>  extern char *get_gdb_completer_word_break_characters (void);
> 
>  extern char *get_gdb_completer_quote_characters (void);
> --- gdb/cli/cli-cmds.c~0        Mon Jul 16 16:33:46 2001
> +++ gdb/cli/cli-cmds.c  Wed Jan  9 08:58:40 2002
> @@ -668,7 +668,8 @@ when gdb is started.", &cmdlist);
>    c->completer = filename_completer;
> 
>    add_com ("quit", class_support, quit_command, "Exit gdb.");
> -  add_com ("help", class_support, help_command, "Print list of commands.");
> +  c = add_com ("help", class_support, help_command, "Print list of commands.");
> +  c->completer = command_completer;
>    add_com_alias ("q", "quit", class_support, 1);
>    add_com_alias ("h", "help", class_support, 1);
> 

-- 
Fernando Nasser
Red Hat Canada Ltd.                     E-Mail:  fnasser@redhat.com
2323 Yonge Street, Suite #300
Toronto, Ontario   M4P 2C9

Re: [RFA] "Help" command and completion

[Michael Snyder]

Eli Zaretskii wrote:
> 
> On Tue, 8 Jan 2002, Michael Snyder wrote:
> 
> > Michael Snyder wrote:
> > >
> > > Eli Zaretskii wrote:
> > > >
> > > > How is no completion better than completion on symbols?
> > >
> > > Because the symbols in the program being debugged
> > > can't possibly have anything to do with gdb commands,
> > > and therefore can't possibly be valid arguments for "help".
> > > Not even if you're debugging gdb.
> >
> > OK to commit?
> 
> Personally, I'd rather have it complete on commands, like you said it
> should.  We already have the function to do that: complete_on_cmdlist,
> so it shouldn't be too hard to make that happen.

Excellent!  I was looking for that, but didn't find it.
My patch was meant to be an incremental improvement, 
yours is the proper way to do things.
 
> In fact, I just took a quick attempt on that; the patch is below.  It
> seems to work for me.  (The diffs are against stock GDB 5.1, since
> that's what I happen to have on the machine I have here.)
> 
> In any case, IIRC completion is Fernando's land, so I guess we need to
> wait for his word.
> 
> 2002-01-09  Eli Zaretskii  <eliz@is.elta.co.il>
> 
>         * completer.c (command_completer): New function.
> 
>         * completer.h <command_completer>: Add prototype.
> 
>         * cli/cli-cmds.c (init_cli_cmds): Make command_completer be the
>         completer for the "help" command.
> 
> --- gdb/completer.c~0   Sun Jul 15 21:17:46 2001
> +++ gdb/completer.c     Wed Jan  9 09:05:48 2002
> @@ -339,6 +339,14 @@ location_completer (char *text, char *wo
>    return list;
>  }
> 
> +/* Complete on command names.  Used by "help".  */
> +char **
> +command_completer (char *text, char *word)
> +{
> +  return complete_on_cmdlist (cmdlist, text, word);
> +}
> +
> +
>  /* Here are some useful test cases for completion.  FIXME: These should
>     be put in the test suite.  They should be tested with both M-? and TAB.
> 
> --- gdb/completer.h~0   Sun Jul 15 21:17:46 2001
> +++ gdb/completer.h     Wed Jan  9 08:52:50 2002
> @@ -29,6 +29,8 @@
> 
>  extern char **location_completer (char *, char *);
> 
> +extern char **command_completer (char *, char *);
> +
>  extern char *get_gdb_completer_word_break_characters (void);
> 
>  extern char *get_gdb_completer_quote_characters (void);
> --- gdb/cli/cli-cmds.c~0        Mon Jul 16 16:33:46 2001
> +++ gdb/cli/cli-cmds.c  Wed Jan  9 08:58:40 2002
> @@ -668,7 +668,8 @@ when gdb is started.", &cmdlist);
>    c->completer = filename_completer;
> 
>    add_com ("quit", class_support, quit_command, "Exit gdb.");
> -  add_com ("help", class_support, help_command, "Print list of commands.");
> +  c = add_com ("help", class_support, help_command, "Print list of commands.");
> +  c->completer = command_completer;
>    add_com_alias ("q", "quit", class_support, 1);
>    add_com_alias ("h", "help", class_support, 1);
>

Re: [RFA] "Help" command and completion

[Michael Snyder]

Fernando Nasser wrote:
> 
> Sorry folks, I just came back from vacations and I still have to catch
> up with all the e-mail.   I saw this one by chance while searching for
> another thing...
> 
> Yes, I agree 100% with you both.
> 
> I suggest that to proper document how we got here, Michael should check
> his patch first and then Eli checks his in.

Nah, my name's in the changelog enough already.
Let's just let Eli check in his superior change.

> 
> Regards to all and Happy New Year.
> 
> Fernando
> 
> Eli Zaretskii wrote:
> >
> > On Tue, 8 Jan 2002, Michael Snyder wrote:
> >
> > > Michael Snyder wrote:
> > > >
> > > > Eli Zaretskii wrote:
> > > > >
> > > > > How is no completion better than completion on symbols?
> > > >
> > > > Because the symbols in the program being debugged
> > > > can't possibly have anything to do with gdb commands,
> > > > and therefore can't possibly be valid arguments for "help".
> > > > Not even if you're debugging gdb.
> > >
> > > OK to commit?
> >
> > Personally, I'd rather have it complete on commands, like you said it
> > should.  We already have the function to do that: complete_on_cmdlist,
> > so it shouldn't be too hard to make that happen.
> >
> > In fact, I just took a quick attempt on that; the patch is below.  It
> > seems to work for me.  (The diffs are against stock GDB 5.1, since
> > that's what I happen to have on the machine I have here.)
> >
> > In any case, IIRC completion is Fernando's land, so I guess we need to
> > wait for his word.
> >
> > 2002-01-09  Eli Zaretskii  <eliz@is.elta.co.il>
> >
> >         * completer.c (command_completer): New function.
> >
> >         * completer.h <command_completer>: Add prototype.
> >
> >         * cli/cli-cmds.c (init_cli_cmds): Make command_completer be the
> >         completer for the "help" command.
> >
> > --- gdb/completer.c~0   Sun Jul 15 21:17:46 2001
> > +++ gdb/completer.c     Wed Jan  9 09:05:48 2002
> > @@ -339,6 +339,14 @@ location_completer (char *text, char *wo
> >    return list;
> >  }
> >
> > +/* Complete on command names.  Used by "help".  */
> > +char **
> > +command_completer (char *text, char *word)
> > +{
> > +  return complete_on_cmdlist (cmdlist, text, word);
> > +}
> > +
> > +
> >  /* Here are some useful test cases for completion.  FIXME: These should
> >     be put in the test suite.  They should be tested with both M-? and TAB.
> >
> > --- gdb/completer.h~0   Sun Jul 15 21:17:46 2001
> > +++ gdb/completer.h     Wed Jan  9 08:52:50 2002
> > @@ -29,6 +29,8 @@
> >
> >  extern char **location_completer (char *, char *);
> >
> > +extern char **command_completer (char *, char *);
> > +
> >  extern char *get_gdb_completer_word_break_characters (void);
> >
> >  extern char *get_gdb_completer_quote_characters (void);
> > --- gdb/cli/cli-cmds.c~0        Mon Jul 16 16:33:46 2001
> > +++ gdb/cli/cli-cmds.c  Wed Jan  9 08:58:40 2002
> > @@ -668,7 +668,8 @@ when gdb is started.", &cmdlist);
> >    c->completer = filename_completer;
> >
> >    add_com ("quit", class_support, quit_command, "Exit gdb.");
> > -  add_com ("help", class_support, help_command, "Print list of commands.");
> > +  c = add_com ("help", class_support, help_command, "Print list of commands.");
> > +  c->completer = command_completer;
> >    add_com_alias ("q", "quit", class_support, 1);
> >    add_com_alias ("h", "help", class_support, 1);
> >
> 
> --
> Fernando Nasser
> Red Hat Canada Ltd.                     E-Mail:  fnasser@redhat.com
> 2323 Yonge Street, Suite #300
> Toronto, Ontario   M4P 2C9

Re: [RFA] "Help" command and completion

[Eli Zaretskii]

> Date: Wed, 09 Jan 2002 10:22:59 -0800
> From: Michael Snyder <msnyder@redhat.com>
> 
> Let's just let Eli check in his superior change.

I tried, but couldn't: neither sourceware.cygnus.com nor
sources.redhat.com won't let me in.  They both say:

  Permission denied (publickey,keyboard-interactive)

What am I missing?  I'm guessing that this is due to the sourceware
being shut down, but I don't think anybody said that I should change
what I type to access the CVS.  Did I miss a message posted here?

Re: [RFA] "Help" command and completion

[Daniel Jacobowitz]

On Wed, Jan 09, 2002 at 09:09:39PM +0200, Eli Zaretskii wrote:
> > Date: Wed, 09 Jan 2002 10:22:59 -0800
> > From: Michael Snyder <msnyder@redhat.com>
> > 
> > Let's just let Eli check in his superior change.
> 
> I tried, but couldn't: neither sourceware.cygnus.com nor
> sources.redhat.com won't let me in.  They both say:
> 
>   Permission denied (publickey,keyboard-interactive)
> 
> What am I missing?  I'm guessing that this is due to the sourceware
> being shut down, but I don't think anybody said that I should change
> what I type to access the CVS.  Did I miss a message posted here?

That looks to me as if it's using SSH2, and sources only has SSH1 keys
on it.  Did you switch your sourceware CVSroot to sources recently?  If
so, adding:
Host sources.redhat.com
  Protocol 1

to .ssh/config may help.

-- 
Daniel Jacobowitz                           Carnegie Mellon University
MontaVista Software                         Debian GNU/Linux Developer

Re: [RFA] "Help" command and completion

[Christopher Faylor]

On Wed, Jan 09, 2002 at 02:22:42PM -0500, Daniel Jacobowitz wrote:
>On Wed, Jan 09, 2002 at 09:09:39PM +0200, Eli Zaretskii wrote:
>> > Date: Wed, 09 Jan 2002 10:22:59 -0800
>> > From: Michael Snyder <msnyder@redhat.com>
>> > 
>> > Let's just let Eli check in his superior change.
>> 
>> I tried, but couldn't: neither sourceware.cygnus.com nor
>> sources.redhat.com won't let me in.  They both say:
>> 
>>   Permission denied (publickey,keyboard-interactive)
>> 
>> What am I missing?  I'm guessing that this is due to the sourceware
>> being shut down, but I don't think anybody said that I should change
>> what I type to access the CVS.  Did I miss a message posted here?
>
>That looks to me as if it's using SSH2, and sources only has SSH1 keys
>on it.  Did you switch your sourceware CVSroot to sources recently?  If
>so, adding:
>Host sources.redhat.com
>  Protocol 1
>
>to .ssh/config may help.

Sounds like a good suggestion.  The only changes I've made to sources.redhat.com
wrt sourceware.cygnus.com is to bounce email to sourceware.cygnus.com.

cgf

Re: [RFA] "Help" command and completion

[Eli Zaretskii]

> Date: Wed, 09 Jan 2002 10:22:59 -0800
> From: Michael Snyder <msnyder@redhat.com>
> 
> Let's just let Eli check in his superior change.

Committed on the trunk.  If someone thinks it is safe enough to go to
the branch as well, please say so.

Re: [RFA] "Help" command and completion

[Eli Zaretskii]

> Date: Wed, 9 Jan 2002 14:22:42 -0500
> From: Daniel Jacobowitz <drow@mvista.com>
> 
> That looks to me as if it's using SSH2, and sources only has SSH1 keys
> on it.  Did you switch your sourceware CVSroot to sources recently?  If
> so, adding:
> Host sources.redhat.com
>   Protocol 1
> 
> to .ssh/config may help.

Thanks, this did the trick.

Re: [RFA] "Help" command and completion

[Eli Zaretskii]

> Date: Wed, 9 Jan 2002 14:24:36 -0500
> From: Christopher Faylor <cgf@redhat.com>
> 
> >Host sources.redhat.com
> >  Protocol 1
> >
> >to .ssh/config may help.
> 
> Sounds like a good suggestion.  The only changes I've made to
> sources.redhat.com wrt sourceware.cygnus.com is to bounce email to
> sourceware.cygnus.com.

Btw, would it be possible for the GDB Web page to show the CVS
command suitable for developers' access to the CVS repository?  Right
now, there's only a command for anon CVS access.

I sometimes have to work from weird locations where I don't have shell
scripts ready with all the magic words, like the exact name of the
server or the root of the CVS tree, and my memory is not what it was
10 years ago...  It's annoying to have to ftp to one of my other
accounts and fetch the shell scripts from there, especially if all I
need to do is one simlpe commit.

For comparison, http://savannah.gnu.org/projects/emacs has clear
instructions for both anon and developer access to the CVS tree,
which helps a lot.

TIA

Re: [RFA] "Help" command and completion

[Andrew Cagney]

> Date: Wed, 9 Jan 2002 14:24:36 -0500
>> From: Christopher Faylor <cgf@redhat.com>
>> 
> 
>> >Host sources.redhat.com
>> > Protocol 1
>> >
>> >to .ssh/config may help.
> 
>> 
>> Sounds like a good suggestion.  The only changes I've made to
>> sources.redhat.com wrt sourceware.cygnus.com is to bounce email to
>> sourceware.cygnus.com.
> 
> 
> Btw, would it be possible for the GDB Web page to show the CVS
> command suitable for developers' access to the CVS repository?  Right
> now, there's only a command for anon CVS access.


Personally I really don't have an opinion.

I suspect a link to a more generic page covering src (with links to 
further CVS info) might be more useful.


> I sometimes have to work from weird locations where I don't have shell
> scripts ready with all the magic words, like the exact name of the
> server or the root of the CVS tree, and my memory is not what it was
> 10 years ago...  It's annoying to have to ftp to one of my other
> accounts and fetch the shell scripts from there, especially if all I
> need to do is one simlpe commit.
> 
> For comparison, http://savannah.gnu.org/projects/emacs has clear
> instructions for both anon and developer access to the CVS tree,
> which helps a lot.


Andrew

Re: [RFA] "Help" command and completion

[Andrew Cagney]

> Btw, would it be possible for the GDB Web page to show the CVS
> command suitable for developers' access to the CVS repository?  Right
> now, there's only a command for anon CVS access.
> 
> 
> Personally I really don't have an opinion.
> 
> I suspect a link to a more generic page covering src (with links to further CVS info) might be more useful.


I ended up with a second yes so I've added a very brief note to 
http://sources.redhat.com/gdb/current/ page.

Andrew

Re: [RFA] "Help" command and completion

[Andrew Cagney]

> Date: Wed, 09 Jan 2002 10:22:59 -0800
>> From: Michael Snyder <msnyder@redhat.com>
>> 
>> Let's just let Eli check in his superior change.
> 
> 
> Committed on the trunk.  If someone thinks it is safe enough to go to
> the branch as well, please say so.


Don't worry, it looks like there'll be a 5.2 soon anyway.

Andrew