-inferior-tty-show and -inferior-tty-set

-inferior-tty-show and -inferior-tty-set

[Nick Roberts]

-inferior-tty-show and -inferior-tty-set seem to accept any number of
arguments:

(gdb) 
-inferior-tty-show asdf
^done
(gdb) 

(gdb) 
-inferior-tty-set asdf qwert
^done
(gdb) 

With mi_cmd_inferior_tty_set there is no test.  mi_cmd_inferior_tty_show
uses mi_valid_noargs which returns 1 even with arguments (mi_getopt
only returns something other than -1 if it finds an option i.e an argument
starting with `-').

Since mi_cmd_file_list_exec_source_file and mi_cmd_file_list_exec_source_files
also use mi_valid_noargs, their associated MI commands also accept any number
of arguments.

Nick

Re: -inferior-tty-show and -inferior-tty-set

[Daniel Jacobowitz]

On Wed, Jul 13, 2005 at 02:58:09PM +1200, Nick Roberts wrote:
> 
> -inferior-tty-show and -inferior-tty-set seem to accept any number of
> arguments:
> 
> (gdb) 
> -inferior-tty-show asdf
> ^done
> (gdb) 
> 
> (gdb) 
> -inferior-tty-set asdf qwert
> ^done
> (gdb) 

These we need to rediscuss whether to have at all - I'm still inclined
to yes, you're still inclined to no, I haven't had time to sit down and
write about it.

> With mi_cmd_inferior_tty_set there is no test.  mi_cmd_inferior_tty_show
> uses mi_valid_noargs which returns 1 even with arguments (mi_getopt
> only returns something other than -1 if it finds an option i.e an argument
> starting with `-').
> 
> Since mi_cmd_file_list_exec_source_file and mi_cmd_file_list_exec_source_files
> also use mi_valid_noargs, their associated MI commands also accept any number
> of arguments.

Ouch.  That definitely ought to be fixed...

-- 
Daniel Jacobowitz
CodeSourcery, LLC

Re: -inferior-tty-show and -inferior-tty-set

[Bob Rossi]

On Wed, Jul 13, 2005 at 02:58:09PM +1200, Nick Roberts wrote:
> 
> -inferior-tty-show and -inferior-tty-set seem to accept any number of
> arguments:
> 
> (gdb) 
> -inferior-tty-show asdf
> ^done
> (gdb) 
> 
> (gdb) 
> -inferior-tty-set asdf qwert
> ^done
> (gdb) 
> 
> With mi_cmd_inferior_tty_set there is no test.  mi_cmd_inferior_tty_show
> uses mi_valid_noargs which returns 1 even with arguments (mi_getopt
> only returns something other than -1 if it finds an option i.e an argument
> starting with `-').

OK. There could be a problem here. Not sure. A valid MI option is,
 mi-command ==>
  [ token ] "-" operation ( " " option )* [ " --" ] ( " " parameter )* nl
basically, arguments to an mi function begin with a '-'. 

So,
   (gdb)
   -file-list-exec-source-file a
   ^done,line="1",file="test.c",fullname="/home/bob/cvs/gdb/original/objdir/gdb/test.c"
   (gdb)
   -file-list-exec-source-file -a
   &"mi_cmd_file_list_exec_source_file: Unknown option ``a''\n"
   ^error,msg="mi_cmd_file_list_exec_source_file: Unknown option ``a''"
   (gdb)

Should we make mi_valid_noargs take care of the case when invalid
arguments are passed to the MI command? or should mi_getopt return an
error? I'll have to think a little about this.

Bob Rossi

Re: -inferior-tty-show and -inferior-tty-set

[Nick Roberts]

 > OK. There could be a problem here. Not sure. A valid MI option is,
 >  mi-command ==>
 >   [ token ] "-" operation ( " " option )* [ " --" ] ( " " parameter )* nl
 > basically, arguments to an mi function begin with a '-'. 
 > 
 > So,
 >    (gdb)
 >    -file-list-exec-source-file a
 >    ^done,line="1",file="test.c",fullname="/home/bob/cvs/gdb/original/objdir/gdb/test.c"
 >    (gdb)
 >    -file-list-exec-source-file -a
 >    &"mi_cmd_file_list_exec_source_file: Unknown option ``a''\n"
 >    ^error,msg="mi_cmd_file_list_exec_source_file: Unknown option ``a''"
 >    (gdb)
 > 
 > Should we make mi_valid_noargs take care of the case when invalid
 > arguments are passed to the MI command? or should mi_getopt return an
 > error? I'll have to think a little about this.

Instead of

  if ( !mi_valid_noargs("mi_cmd_file_list_exec_source_file", argc, argv) )
    error (_("mi_cmd_file_list_exec_source_file: Usage: No args"));

why not just use:

  if (argc > 0)
    error (_("mi_cmd_file_list_exec_source_file: Usage: No args"));

Incidentally there is no mention of mi_valid_noargs in the ChangeLogs.  If
the above suggestion works, perhaps its not needed and it could be removed.

Nick

Re: -inferior-tty-show and -inferior-tty-set

[Bob Rossi]

On Thu, Jul 14, 2005 at 10:57:13AM +1200, Nick Roberts wrote:
>  > OK. There could be a problem here. Not sure. A valid MI option is,
>  >  mi-command ==>
>  >   [ token ] "-" operation ( " " option )* [ " --" ] ( " " parameter )* nl
>  > basically, arguments to an mi function begin with a '-'. 
>  > 
>  > So,
>  >    (gdb)
>  >    -file-list-exec-source-file a
>  >    ^done,line="1",file="test.c",fullname="/home/bob/cvs/gdb/original/objdir/gdb/test.c"
>  >    (gdb)
>  >    -file-list-exec-source-file -a
>  >    &"mi_cmd_file_list_exec_source_file: Unknown option ``a''\n"
>  >    ^error,msg="mi_cmd_file_list_exec_source_file: Unknown option ``a''"
>  >    (gdb)
>  > 
>  > Should we make mi_valid_noargs take care of the case when invalid
>  > arguments are passed to the MI command? or should mi_getopt return an
>  > error? I'll have to think a little about this.
> 
> Instead of
> 
>   if ( !mi_valid_noargs("mi_cmd_file_list_exec_source_file", argc, argv) )
>     error (_("mi_cmd_file_list_exec_source_file: Usage: No args"));
> 
> why not just use:
> 
>   if (argc > 0)
>     error (_("mi_cmd_file_list_exec_source_file: Usage: No args"));
> 
> Incidentally there is no mention of mi_valid_noargs in the ChangeLogs.  If
> the above suggestion works, perhaps its not needed and it could be removed.

Hmmm, I added the mi_valid_noargs as one of my first GDB commits. Don't
know how it didn't get in the change log. 

I think everywhere that does "if (argc >0)" is broke. I could be wrong
about this though. Here is the reason that Andrew asked me to create
this function,
    http://sourceware.org/ml/gdb-patches/2003-04/msg00030.html

Thanks,
Bob Rossi

Re: -inferior-tty-show and -inferior-tty-set

[Nick Roberts]

 > I think everywhere that does "if (argc >0)" is broke. I could be wrong
 > about this though. Here is the reason that Andrew asked me to create
 > this function,
 >     http://sourceware.org/ml/gdb-patches/2003-04/msg00030.html

It looks to me that the intention is to make:

-file-list-exec-source-file --

legitimate.  Perhaps we could use the condition"(argc > 0)" for now and worry
about such pathological cases later, when we could also improve mi_getopt to
other things like long options (as Daniel has suggested).

In fact I would like to add an argument to -file-list-exec-source-file to
find the fullname of a file i.e

(gdb)
-file-list-exec-source-file myprint.c
^done,line="16",file="myprint.c",fullname="/home/nick/myprint.c"
(gdb)

where myprog.c is not necessarily the current source file.  At the moment, I
do:

(gdb) 
list myprint.c:1
&"list myprint.c:1\n"
~"1\t/* -*- compile-command: \"cc -g3 -c myprint.c\"; -*-  */\n"
~"2\t\n"
~"3\tint i22;\n"
~"4\t\n"
~"5\tstruct values;\n"
~"6\tstruct values1;\n"
~"7\t\n"
~"8\t#define GOTCHA 9\n"
~"9\t\n"
~"10\tmyprint (int i,int j) {\n"
^done
(gdb) 
-file-list-exec-source-file
^done,line="11",file="myprint.c",fullname="/home/nick/myprint.c"
(gdb) 

to make the file that I'm interested in current first (I'm thinking of a
situation where user the adds to the search path for source files *after* the
start of the debug session).

Nick