Re: [RFA] Add new cmd line parameter "--pid" for attach.

[Michael Snyder <msnyder@redhat.com>]

Elena Zannoni wrote:
> 
> Michael Snyder writes:
>  >
>  > Currently if you invoke gdb as:
>  >
>  >      gdb filename 12345
>  >
>  > gdb will attempt to open a corefile called "12345", and if that
>  > fails it will print a "file not found" warning, and then attempt
>  > to attach to a process "12345".
>  >
>  > There is a "--core <filename>" command-line argument,
>  > so that you can specify a corefile without a symbol file:
>  >
>  >      gdb --core <filename>
>  >
>  > but there is no "--pid" option to allow you to specify
>  > a process-id without a symbol file.
>  >
>  > This patch does two things:
>  >
>  > 1) Add a "--pid" option to allow specification of an attach pid.
>  >
> 
> This bit is approved.
> 
>  > 2) If the second argument (after the symbol-file) begins with
>  > a digit, try attach first instead of trying to open it as a
>  > corefile first.  This eliminates the "file not found" warning.
>  >
> 
> About this, I have a question, what happens if you have a corefile
> whose name starts with a digit? I tried it and I get an error:
> 
> [ezannoni@localhost gdb]$ ./gdb -nw ./gdb 2222core
> GNU gdb 2002-01-03-cvs
> Copyright 2001 Free Software Foundation, Inc.
> GDB is free software, covered by the GNU General Public License, and you are
> welcome to change it and/or distribute copies of it under certain conditions.
> Type "show copying" to see the conditions.
> There is absolutely no warranty for GDB.  Type "show warranty" for details.
> This GDB was configured as "i686-pc-linux-gnu"...
> Attaching to program: /home/ezannoni/sources/native/gdb/gdb, process 2222
> ptrace: No such process.
> 
> then it proceeds normally to figure out it's a core file.

Right -- this is actually the reverse of the old behavior
(before my change).  Previously the algorythm was this:

        Try to open a corefile
        on failure, if isdigit(string[0])
                try to attach a pid.

So if it was really a pid, you always got an error when
it tried to open it as a corefile.  Now the algorythm is:

        if isdigit (string[0])
                try to attach a pid
                on failure, try to open a corefile
        else try to open a corefile

So the only time you will get a failure warning is
if you have a corefile whose name begins with a digit.
I think that's an improvement (warning should be less
frequent). 

> 
> I can be convinced that a digit is more likely to indicate a pid than
> a corefile, but would there be a way to make that error be silent? I
> realize that those error messages are generated in the bowels of gdb,
> and it may be really hard to fix that (gee, isn't this something the
> insight people have some opinion about? :-)
> 
> What's that bit about input-radix and output-radix?

Ummm... D'oh!  Sorry, that doesn't belong there.
Please ignore it, I'll strip it out before committing.

> 
> Elena
> 
>  > 2002-01-04  Michael Snyder  <msnyder@redhat.com>
>  >
>  >      * main.c (captured_main): Add new command line option "--pid".
>  >      If the second command line argument (following the symbol-file)
>  >      begins with a digit, try to attach to it before trying to open
>  >      it as a corefile.
>  >      (print_gdb_help): Document the "--pid" argument.
>  >
>  > Index: main.c
>  > ===================================================================
>  > RCS file: /cvs/src/src/gdb/main.c,v
>  > retrieving revision 1.14
>  > diff -c -3 -p -r1.14 main.c
>  > *** main.c   2001/11/22 00:23:12     1.14
>  > --- main.c   2002/01/05 03:15:43
>  > *************** captured_main (void *data)
>  > *** 239,244 ****
>  > --- 239,246 ----
>  >         {"e", required_argument, 0, 'e'},
>  >         {"core", required_argument, 0, 'c'},
>  >         {"c", required_argument, 0, 'c'},
>  > +       {"pid", required_argument, 0, 'p'},
>  > +       {"p", required_argument, 0, 'p'},
>  >         {"command", required_argument, 0, 'x'},
>  >         {"version", no_argument, &print_version, 1},
>  >         {"x", required_argument, 0, 'x'},
>  > *************** captured_main (void *data)
>  > *** 320,325 ****
>  > --- 322,331 ----
>  >        case 'c':
>  >          corearg = optarg;
>  >          break;
>  > +      case 'p':
>  > +        /* "corearg" is shared by "--core" and "--pid" */
>  > +        corearg = optarg;
>  > +        break;
>  >        case 'x':
>  >          cmdarg[ncmd++] = optarg;
>  >          if (ncmd >= cmdsize)
>  > *************** extern int gdbtk_test (char *);
>  > *** 463,470 ****
>  >            execarg = argv[optind];
>  >            break;
>  >          case 2:
>  > !          /* FIXME: The documentation says this can be a
>  > !             "ProcID". as well.  */
>  >            corearg = argv[optind];
>  >            break;
>  >          case 3:
>  > --- 469,476 ----
>  >            execarg = argv[optind];
>  >            break;
>  >          case 2:
>  > !          /* The documentation says this can be a "ProcID" as well.
>  > !             We will try it as both a corefile and a pid.  */
>  >            corearg = argv[optind];
>  >            break;
>  >          case 3:
>  > *************** extern int gdbtk_test (char *);
>  > *** 586,597 ****
>  >
>  >     if (corearg != NULL)
>  >       {
>  > !       if (catch_command_errors (core_file_command, corearg, !batch,
RETURN_MASK_ALL) == 0)
>  >      {
>  > !      /* See if the core file is really a PID. */
>  > !      if (isdigit (corearg[0]))
>  > !        catch_command_errors (attach_command, corearg, !batch,
RETURN_MASK_ALL);
>  >      }
>  >       }
>  >
>  >     if (ttyarg != NULL)
>  > --- 592,611 ----
>  >
>  >     if (corearg != NULL)
>  >       {
>  > !       /* corearg may be either a corefile or a pid.
>  > !     If its first character is a digit, try attach first
>  > !     and then corefile.  Otherwise try corefile first. */
>  > !
>  > !       if (isdigit (corearg[0]))
>  >      {
>  > !      if (catch_command_errors (attach_command, corearg,
>  > !                                !batch, RETURN_MASK_ALL) == 0)
>  > !        catch_command_errors (core_file_command, corearg,
>  > !                              !batch, RETURN_MASK_ALL);
>  >      }
>  > +       else /* Can't be a pid, better be a corefile. */
>  > +    catch_command_errors (core_file_command, corearg,
>  > +                          !batch, RETURN_MASK_ALL);
>  >       }
>  >
>  >     if (ttyarg != NULL)
>  > *************** extern int gdbtk_test (char *);
>  > *** 616,621 ****
>  > --- 630,646 ----
>  >      catch_command_errors (source_command, gdbinit, 0, RETURN_MASK_ALL);
>  >         }
>  >
>  > +   /* These need to be set this late in the initialization to ensure that
>  > +      they are defined for the current environment.  They define the
>  > +      radix variables needed by a save-breakpoints file to preserve the
>  > +      radix across the breakpoints restoration assuming they are restored
>  > +      using the -x (-command) command line options.  */
>  > +
>  > +   set_internalvar (lookup_internalvar ("input_radix"),
>  > +               value_from_longest (builtin_type_int, (LONGEST)
input_radix));
>  > +   set_internalvar (lookup_internalvar ("output_radix"),
>  > +               value_from_longest (builtin_type_int, (LONGEST)
output_radix));
>  > +
>  >     for (i = 0; i < ncmd; i++)
>  >       {
>  >   #if 0
>  > *************** Options:\n\n\
>  > *** 752,757 ****
>  > --- 777,783 ----
>  >     --cd=DIR           Change current directory to DIR.\n\
>  >     --command=FILE     Execute GDB commands from FILE.\n\
>  >     --core=COREFILE    Analyze the core dump COREFILE.\n\
>  > +   --pid=PID          Attach to running process PID.\n\
>  >   ", stream);
>  >     fputs_unfiltered ("\
>  >     --dbx              DBX compatibility mode.\n\