Re: [patch] Fix gdbserver vRun parsing again

Re: [patch] Fix gdbserver vRun parsing again

[Daniel Jacobowitz]

On Mon, Nov 24, 2008 at 05:59:36PM +0000, Pedro Alves wrote:
> I tried it out, and it does do the right thing, and I
> also confirmed the original crash isn't reintroduced.

Thanks!  I checked it in.

-- 
Daniel Jacobowitz
CodeSourcery

Re: [patch] Fix gdbserver vRun parsing again

[Pedro Alves]

On Monday 24 November 2008 17:31:35, Daniel Jacobowitz wrote:
> As part of closing PR 2474, I took another look at Pedro's fix last
> month to vRun.  I approved it at the time but now I can see that he
> and I were both a little confused about the documented semantics :-(
> With Pedro's patch, "target extended-remote", "run one", "run two"
> will never run the program with argv[1] == "two".  We're always
> supposed to use the new argv.  "vRun;" means "old program, no args";
> "vRun;;XXXX" means "old program, decoded arg XXXX".
> 

Oh, very sorry about that.  I should have sticked to fixing
the crash.  :-(

> This patch seems to do the right thing; Pedro, want to take a look at
> this obviously tricky code before I change it?

I tried it out, and it does do the right thing, and I
also confirmed the original crash isn't reintroduced.

Thanks.

-- 
Pedro Alves

[patch] Fix gdbserver vRun parsing again

[Daniel Jacobowitz]

As part of closing PR 2474, I took another look at Pedro's fix last
month to vRun.  I approved it at the time but now I can see that he
and I were both a little confused about the documented semantics :-(

With Pedro's patch, "target extended-remote", "run one", "run two"
will never run the program with argv[1] == "two".  We're always
supposed to use the new argv.  "vRun;" means "old program, no args";
"vRun;;XXXX" means "old program, decoded arg XXXX".

This patch seems to do the right thing; Pedro, want to take a look at
this obviously tricky code before I change it?

-- 
Daniel Jacobowitz
CodeSourcery

2008-11-24  Daniel Jacobowitz  <dan@codesourcery.com>

	* server.c (handle_v_run): Always use the supplied argument list.

Index: server.c
===================================================================
RCS file: /cvs/src/src/gdb/gdbserver/server.c,v
retrieving revision 1.79
diff -u -p -r1.79 server.c
--- server.c	10 Oct 2008 14:06:05 -0000	1.79
+++ server.c	24 Nov 2008 17:24:44 -0000
@@ -1071,7 +1071,7 @@ handle_v_run (char *own_buf, char *statu
       new_argc++;
     }
 
-  new_argv = malloc ((new_argc + 2) * sizeof (char *));
+  new_argv = calloc (new_argc + 2, sizeof (char *));
   i = 0;
   for (p = own_buf + strlen ("vRun;"); *p; p = next_p)
     {
@@ -1096,10 +1096,8 @@ handle_v_run (char *own_buf, char *statu
 
   if (new_argv[0] == NULL)
     {
-      /* GDB didn't specify a program to run.  Try to use the argv
-	 from the last run: either from the last vRun with a non-empty
-	 argv, or from what the user specified if gdbserver was
-	 started as: `gdbserver :1234 PROG ARGS'.  */
+      /* GDB didn't specify a program to run.  Use the program from the
+	 last run with the new argument list.  */
 
       if (program_argv == NULL)
 	{
@@ -1107,20 +1105,17 @@ handle_v_run (char *own_buf, char *statu
 	  return 0;
 	}
 
-      /* We can reuse the old args.  We don't need this then.  */
-      free (new_argv);
+      new_argv[0] = strdup (program_argv[0]);
     }
-  else
+
+  /* Free the old argv.  */
+  if (program_argv)
     {
-      /* Free the old argv.  */
-      if (program_argv)
-	{
-	  for (pp = program_argv; *pp != NULL; pp++)
-	    free (*pp);
-	  free (program_argv);
-	}
-      program_argv = new_argv;
+      for (pp = program_argv; *pp != NULL; pp++)
+	free (*pp);
+      free (program_argv);
     }
+  program_argv = new_argv;
 
   *signal = start_inferior (program_argv, status);
   if (*status == 'T')