The joy of $0. Undoing the #16399 makes Andreas'

tests (see [perl #22811]) pass (yes, padding with space	instead
of nul makes no sense, but that seems to work, maybe Linux does
some deep magic in ps(1)?); moving the PL_origalen computation
earlier makes also the threaded-first case fully pass.

But in general modifying the argv[] is very non-portable.
(e.g. in Tru64 it seems to be limited to the size of the
original argv[0] since the argv[] are not contiguous?)

Everybody should just have setproctitle().

p4raw-id: //depot/perl@19884

3 files changed, 67 insertions, 57 deletions

diff --git a/mg.c b/mg.c
index 98ccb34e3d..8182376662 100644
--- a/mg.c
+++ b/mg.c
@@ -2372,60 +2372,32 @@ Perl_magic_set(pTHX_ SV *sv, MAGIC *mg)
 	     pstat(PSTAT_SETCMD, un, len, 0, 0);
 	}
 #endif
-	if (!PL_origalen) {
-	    s = PL_origargv[0];
-	    s += strlen(s);
-	    /* See if all the arguments are contiguous in memory */
-	    for (i = 1; i < PL_origargc; i++) {
-		if (PL_origargv[i] == s + 1
-#ifdef OS2
-		    || PL_origargv[i] == s + 2
-#endif
-		   )
-		{
-		    ++s;
-		    s += strlen(s);	/* this one is ok too */
-		}
-		else
-		    break;
-	    }
-	    /* can grab env area too? */
-	    if (PL_origenviron
-#ifdef USE_ITHREADS
-		&& PL_curinterp == aTHX
-#endif
-		&& (PL_origenviron[0] == s + 1))
-	    {
-		my_setenv("NoNe  SuCh", Nullch);
-					    /* force copy of environment */
-		for (i = 0; PL_origenviron[i]; i++)
-		    if (PL_origenviron[i] == s + 1) {
-			++s;
-			s += strlen(s);
-		    }
-		    else
-			break;
-	    }
-	    PL_origalen = s - PL_origargv[0];
-	}
+	/* PL_origalen is set in perl_parse() to be the sum
+	 * of the contiguous argv[] elements plus the size of
+	 * the env in case that is contiguous with the argv[].
+	 *
+	 * This means that in the worst case the area we are able
+	 * to modify is limited to the size of the original argv[0].
+	 * --jhi */
 	s = SvPV_force(sv,len);
-	i = len;
-	if (i >= (I32)PL_origalen) {
-	    i = PL_origalen;
-	    /* don't allow system to limit $0 seen by script */
-	    /* SvCUR_set(sv, i); *SvEND(sv) = '\0'; */
-	    Copy(s, PL_origargv[0], i, char);
-	    s = PL_origargv[0]+i;
-	    *s = '\0';
+	if (len >= (I32)PL_origalen) {
+	    /* Longer than original, will be truncated. */
+	    Copy(s, PL_origargv[0], PL_origalen, char);
+	    PL_origargv[0][PL_origalen - 1] = 0;
 	}
 	else {
-	    Copy(s, PL_origargv[0], i, char);
-	    s = PL_origargv[0]+i;
-	    *s++ = '\0';
-	    while (++i < (I32)PL_origalen)
-		*s++ = '\0';
+	    /* Shorter than original, will be padded. */
+	    Copy(s, PL_origargv[0], len, char);
+	    PL_origargv[0][len] = 0;
+	    memset(PL_origargv[0] + len + 1,
+		   /* Is the space counterintuitive?  Yes.
+		    * (You were expecting \0?)  
+		    * Does it work?  Seems to.  (In Linux at least.)
+		    * --jhi */
+		   (int)' ',
+		   PL_origalen - len - 1);
 	    for (i = 1; i < PL_origargc; i++)
-		PL_origargv[i] = Nullch;
+		 PL_origargv[i] = 0;
 	}
 	UNLOCK_DOLLARZERO_MUTEX;
 	break;
diff --git a/perl.c b/perl.c
index bb45684638..2f07f4cb80 100644
--- a/perl.c
+++ b/perl.c
@@ -933,6 +933,43 @@ setuid perl scripts securely.\n");
     PL_origargc = argc;
     PL_origargv = argv;
 
+    {
+	 char *s = PL_origargv[0];
+	 int i;
+
+	 s += strlen(s);
+	 /* See if all the arguments are contiguous in memory */
+	 for (i = 1; i < PL_origargc; i++) {
+	      if (PL_origargv[i] == s + 1
+#ifdef OS2
+		  || PL_origargv[i] == s + 2
+#endif
+		   )
+	      {
+		   ++s;
+		   s += strlen(s);	/* this one is ok too */
+	      }
+	      else
+		   break;
+	 }
+	 /* Can we grab env area too to be used as the area for $0
+	  * (in case we later modify it)? */
+	 if (PL_origenviron
+	     && (PL_origenviron[0] == s + 1))
+	 {
+	      my_setenv("NoNe  SuCh", Nullch);
+	      /* force copy of environment */
+	      for (i = 0; PL_origenviron[i]; i++)
+		   if (PL_origenviron[i] == s + 1) {
+			++s;
+			s += strlen(s);
+		   }
+		   else
+			break;
+	 }
+	 PL_origalen = s - PL_origargv[0];
+    }
+
     if (PL_do_undump) {
 
 	/* Come here if running an undumped a.out. */
diff --git a/pod/perlvar.pod b/pod/perlvar.pod
index ad791dd71b..7667f1d75d 100644
--- a/pod/perlvar.pod
+++ b/pod/perlvar.pod
@@ -838,16 +838,17 @@ and C<$)> can be swapped only on machines supporting setregid().
 
 =item $0
 
-Contains the name of the program being executed.  On some operating
-systems assigning to C<$0> modifies the argument area that the B<ps>
-program sees.  This is more useful as a way of indicating the current
-program state than it is for hiding the program you're running.
-(Mnemonic: same as B<sh> and B<ksh>.)
+Contains the name of the program being executed.  On some (read: not
+all) operating systems assigning to C<$0> modifies the argument area
+that the B<ps> program sees.  Also note that depending on the platform,
+the maximum length of C<$0> may be limited to the space occupied by
+the original C<$0>.  This is more useful as a way of indicating the
+current program state than it is for hiding the program you're
+running.  (Mnemonic: same as B<sh> and B<ksh>.)
 
 Note for BSD users: setting C<$0> does not completely remove "perl"
 from the ps(1) output.  For example, setting C<$0> to C<"foobar"> will
-result in C<"perl: foobar (perl)">.  This is an operating system
-feature.
+result in C<"perl: foobar (perl)">.  This is an operating system feature.
 
 In multithreaded scripts Perl coordinates the threads so that any
 thread may modify its copy of the C<$0> and the change becomes visible