[PATCH] Use multi-arch'd START_INFERIOR_TRAPS_EXPECTED on Alpha target

[PATCH] Use multi-arch'd START_INFERIOR_TRAPS_EXPECTED on Alpha target

[Jason R Thorpe]

[-- Attachment #1: Type: text/plain, Size: 764 bytes --]

The following can happen when my STARTUP_WITH_SHELL and
START_INFERIOR_TRAPS_EXPECTED change to gdbarch go in.

Note that config/alpha/tm-fbsd.h is now empty, and thus deleted.  The
same also goes for config/alpha/tm-nbsd.h in my patch that adds the
alpha-*-netbsd* target.

	* Makefile.in (ALLDEPFILES): Add alpha-osf1-tdep.c.
	(alpha-osf1-tdep.o): New dependency list.
	* alpha-osf1-tdep.c: New file.
	* config/alpha/alpha-osf1.mt (TDEPFILES): Add alpha-osf1-tdep.o.
	* config/alpha/fbsd.mt (TM_FILE): Set to tm-alpha.h.
	* config/alpha/tm-alpha.h (START_INFERIOR_TRAPS_EXPECTED): Remove.
	* config/alpha/tm-alphalinux.h (START_INFERIOR_TRAPS_EXPECTED): Remove.
	* config/alpha/tm-fbsd.h: Delete file.

-- 
        -- Jason R. Thorpe <thorpej@wasabisystems.com>

[-- Attachment #2: alpha-start-patch --]
[-- Type: text/plain, Size: 6361 bytes --]

Index: Makefile.in
===================================================================
RCS file: /cvs/src/src/gdb/Makefile.in,v
retrieving revision 1.172
diff -u -r1.172 Makefile.in
--- Makefile.in	21 Apr 2002 22:03:10 -0000	1.172
+++ Makefile.in	22 Apr 2002 01:06:26 -0000
@@ -1174,8 +1174,8 @@
 ALLDEPFILES = 29k-share/udi/udip2soc.c 29k-share/udi/udr.c \
 	29k-share/udi/udi2go32.c \
 	a29k-tdep.c a68v-nat.c \
-	alpha-nat.c alphabsd-nat.c alpha-tdep.c alpha-linux-tdep.c \
-	alphafbsd-tdep.c \
+	alpha-nat.c alphabsd-nat.c alpha-tdep.c \
+	alpha-linux-tdep.c alpha-osf1-tdep.c alphafbsd-tdep.c \
 	arm-linux-nat.c arm-linux-tdep.c arm-tdep.c \
 	armnbsd-nat.c armnbsd-tdep.c \
 	coff-solib.c \
@@ -1250,6 +1250,8 @@
 	$(regcache_h) $(doublest_h) $(BFD_SRC)/elf-bfd.h alpha-tdep.h
 
 alpha-linux-tdep.o: alpha-linux-tdep.c $(defs_h) $(value_h) alpha-tdep.h
+
+alpha-osf1-tdep.o: alpha-osf1-tdep.c $(defs_h) $(value_h) alpha-tdep.h
 
 alphafbsd-tdep.o: alphafbsd-tdep.c $(defs_h) $(value_h) alpha-tdep.h
 
Index: alpha-osf1-tdep.c
===================================================================
RCS file: alpha-osf1-tdep.c
diff -N alpha-osf1-tdep.c
--- /dev/null	1 Jan 1970 00:00:00 -0000
+++ alpha-osf1-tdep.c	22 Apr 2002 01:06:27 -0000
@@ -0,0 +1,37 @@
+/* Target-dependent code for OSF/1 on Alpha.
+   Copyright 2002 Free Software Foundation, Inc.
+
+   This file is part of GDB.
+
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 2 of the License, or
+   (at your option) any later version.
+
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+
+   You should have received a copy of the GNU General Public License
+   along with this program; if not, write to the Free Software
+   Foundation, Inc., 59 Temple Place - Suite 330,
+   Boston, MA 02111-1307, USA.  */
+
+#include "defs.h"
+#include "value.h"
+
+#include "alpha-tdep.h"
+
+static void
+alpha_osf1_init_abi (struct gdbarch_info info,
+                     struct gdbarch *gdbarch)
+{
+  set_gdbarch_start_inferior_traps_expected (gdbarch, 3);
+}
+
+void
+_initialize_alpha_osf1_tdep (void)
+{
+  alpha_gdbarch_register_os_abi (ALPHA_ABI_OSF1, alpha_osf1_init_abi);
+}
Index: config/alpha/alpha-osf1.mt
===================================================================
RCS file: /cvs/src/src/gdb/config/alpha/alpha-osf1.mt,v
retrieving revision 1.1.1.1
diff -u -r1.1.1.1 alpha-osf1.mt
--- config/alpha/alpha-osf1.mt	16 Apr 1999 01:34:14 -0000	1.1.1.1
+++ config/alpha/alpha-osf1.mt	22 Apr 2002 01:06:41 -0000
@@ -1,3 +1,3 @@
 # Target: Little-endian Alpha
-TDEPFILES= alpha-tdep.o
+TDEPFILES= alpha-tdep.o alpha-osf1-tdep.o
 TM_FILE= tm-alpha.h
Index: config/alpha/fbsd.mt
===================================================================
RCS file: /cvs/src/src/gdb/config/alpha/fbsd.mt,v
retrieving revision 1.2
diff -u -r1.2 fbsd.mt
--- config/alpha/fbsd.mt	25 Jan 2001 18:41:23 -0000	1.2
+++ config/alpha/fbsd.mt	22 Apr 2002 01:06:41 -0000
@@ -1,3 +1,3 @@
 # Target: FreeBSD/Alpha
 TDEPFILES= alpha-tdep.o alphafbsd-tdep.o
-TM_FILE= tm-fbsd.h
+TM_FILE= tm-alpha.h
Index: config/alpha/tm-alpha.h
===================================================================
RCS file: /cvs/src/src/gdb/config/alpha/tm-alpha.h,v
retrieving revision 1.22
diff -u -r1.22 tm-alpha.h
--- config/alpha/tm-alpha.h	21 Apr 2002 20:27:29 -0000	1.22
+++ config/alpha/tm-alpha.h	22 Apr 2002 01:06:41 -0000
@@ -36,12 +36,6 @@
 struct value;
 struct symbol;
 
-/* Number of traps that happen between exec'ing the shell 
- * to run an inferior, and when we finally get to 
- * the inferior code.  This is 2 on most implementations.
- */
-#define START_INFERIOR_TRAPS_EXPECTED 3
-
 /* Offset from address of function to start of its code.
    Zero on most machines.  */
 
Index: config/alpha/tm-alphalinux.h
===================================================================
RCS file: /cvs/src/src/gdb/config/alpha/tm-alphalinux.h,v
retrieving revision 1.4
diff -u -r1.4 tm-alphalinux.h
--- config/alpha/tm-alphalinux.h	24 Feb 2002 22:56:04 -0000	1.4
+++ config/alpha/tm-alphalinux.h	22 Apr 2002 01:06:41 -0000
@@ -39,13 +39,6 @@
 #define SIGTRAMP_END(pc)	(SIGTRAMP_START(pc) + 3*4)
 
 
-/* Number of traps that happen between exec'ing the shell to run an
-   inferior, and when we finally get to the inferior code.  This is 2
-   on GNU/Linux and most implementations.  */
-
-#undef START_INFERIOR_TRAPS_EXPECTED
-#define START_INFERIOR_TRAPS_EXPECTED 2
-
 /* Return TRUE if procedure descriptor PROC is a procedure descriptor
    that refers to a dynamically generated sigtramp function.  */
 
Index: config/alpha/tm-fbsd.h
===================================================================
RCS file: config/alpha/tm-fbsd.h
diff -N config/alpha/tm-fbsd.h
--- config/alpha/tm-fbsd.h	21 Apr 2002 20:17:28 -0000	1.5
+++ /dev/null	1 Jan 1970 00:00:00 -0000
@@ -1,32 +0,0 @@
-/* Target-dependent definitions for FreeBSD/Alpha.
-   Copyright 2000, 2001, 2002 Free Software Foundation, Inc.
-
-   This file is part of GDB.
-
-   This program is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published by
-   the Free Software Foundation; either version 2 of the License, or
-   (at your option) any later version.
-
-   This program is distributed in the hope that it will be useful,
-   but WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-   GNU General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program; if not, write to the Free Software
-   Foundation, Inc., 59 Temple Place - Suite 330,
-   Boston, MA 02111-1307, USA.  */
-
-#ifndef TM_FBSD_H
-#define TM_FBSD_H
-
-#include "alpha/tm-alpha.h"
-
-/* Number of traps that happen between exec'ing the shell to run an
-   inferior, and when we finally get to the inferior code.  The
-   default is right for FreeBSD.  */
-
-#undef START_INFERIOR_TRAPS_EXPECTED
-
-#endif /* TM_FBSD_H */

Re: [PATCH] Use multi-arch'd START_INFERIOR_TRAPS_EXPECTED on Alpha target

[Andrew Cagney]

> The following puts STARTUP_WITH_SHELL

Hmm, HP merge.  An intermediate version looked like:

   /* If STARTUP_WITH_SHELL is set, GDB's "run"
!  * will attempts to start up the debugee under a shell.
!  * This is in order for argument-expansion to occur. E.g.,
!  * (gdb) run *
!  * The "*" gets expanded by the shell into a list of files.
!  * While this is a nice feature, it turns out to interact badly
!  * with some of the catch-fork/catch-exec features we have added.
!  * In particular, if the shell does any fork/exec's before
!  * the exec of the target program, that can confuse GDB.
!  * To disable this feature, set STARTUP_WITH_SHELL to 0.
!  * To enable this feature, set STARTUP_WITH_SHELL to 1.
!  * The catch-exec traps expected during start-up will
!  * be 1 if target is not started up with a shell, 2 if it is.
!  * - RT
!  */
   #define STARTUP_WITH_SHELL 1
- #define START_INFERIOR_TRAPS_EXPECTED (STARTUP_WITH_SHELL + 1)

Does STARTUP_WITH_SHELL need to be multi-arched?

I'm wondering if it would be better to make it a variable (``set 
startup-with-shell <boolean>'').  Looking at its uses it appears that 
fork-child.c:startup_inferior() would still work (if it did previously).

Andrew

Re: [PATCH] Use multi-arch'd START_INFERIOR_TRAPS_EXPECTED on Alpha target

[Jason R Thorpe]

On Wed, Apr 24, 2002 at 08:06:31PM -0400, Andrew Cagney wrote:

 > Does STARTUP_WITH_SHELL need to be multi-arched?
 > 
 > I'm wondering if it would be better to make it a variable (``set 
 > startup-with-shell <boolean>'').  Looking at its uses it appears that 
 > fork-child.c:startup_inferior() would still work (if it did previously).

Note that if you set STARTUP_WITH_SHELL to 0, you have to decrement
start_inferior_traps_expected.

So, in all the places that set start_inferior_traps_expected, I guess you'd
have to check:

	set_gdbarch_start_inferior_traps_expected (gdbarch,
	    startup_with_shell ? X : X - 1);

...and then update them all whenever startup_with_shell changed?  Oof, this
is starting to make my head spin :-)

-- 
        -- Jason R. Thorpe <thorpej@wasabisystems.com>

Re: [PATCH] Use multi-arch'd START_INFERIOR_TRAPS_EXPECTED on Alpha target

[Jason R Thorpe]

On Wed, Apr 24, 2002 at 08:06:31PM -0400, Andrew Cagney wrote:

 >    #define STARTUP_WITH_SHELL 1
 > - #define START_INFERIOR_TRAPS_EXPECTED (STARTUP_WITH_SHELL + 1)
 > 
 > Does STARTUP_WITH_SHELL need to be multi-arched?
 > 
 > I'm wondering if it would be better to make it a variable (``set 
 > startup-with-shell <boolean>'').  Looking at its uses it appears that 
 > fork-child.c:startup_inferior() would still work (if it did previously).

Hm.  Actually, it occurs to me that this is more like what we want, but
not quite.  What we really want is:

	STARTUP_WITH_SHELL_ADDS_TRAPS

So, then we get:

#define START_INFERIOR_TRAPS_EXPECTED \
  (startup_with_shell ? (STARTUP_WITH_SHELL_ADDS_TRAPS + 1) : 1)

I.e. the issue is really how many extra traps the shell adds.

-- 
        -- Jason R. Thorpe <thorpej@wasabisystems.com>

Re: [PATCH] Use multi-arch'd START_INFERIOR_TRAPS_EXPECTED on Alpha target

[Andrew Cagney]

> On Wed, Apr 24, 2002 at 08:06:31PM -0400, Andrew Cagney wrote:
> 
>  > #define STARTUP_WITH_SHELL 1
>  > - #define START_INFERIOR_TRAPS_EXPECTED (STARTUP_WITH_SHELL + 1)
>  > > Does STARTUP_WITH_SHELL need to be multi-arched?
>  > > I'm wondering if it would be better to make it a variable (``set 
>  > startup-with-shell <boolean>'').  Looking at its uses it appears that 
>  > fork-child.c:startup_inferior() would still work (if it did previously).

Hmm, ``HP Merge'' ~= ``there be dragons''.

> Hm.  Actually, it occurs to me that this is more like what we want, but
> not quite.  What we really want is:
> 
> 	STARTUP_WITH_SHELL_ADDS_TRAPS
> 
> So, then we get:
> 
> #define START_INFERIOR_TRAPS_EXPECTED \
>   (startup_with_shell ? (STARTUP_WITH_SHELL_ADDS_TRAPS + 1) : 1)
> 
> I.e. the issue is really how many extra traps the shell adds.

I've included a few more excerts.  I think this ``feature'' (only ever 
worked on HP/UX.  Can I suggest taking a step back and making the 
feature work :-)

	Andrew

PS:  I think the relevant bits are:

inftarg.c:
    startup_inferior (START_INFERIOR_TRAPS_EXPECTED);
fork-child.c:

   if (STARTUP_WITH_SHELL)
     inferior_ignoring_startup_exec_events = ntraps;
   else
     inferior_ignoring_startup_exec_events = 0;
   inferior_ignoring_leading_exec_events =
     target_reported_exec_events_per_exec_call () - 1;

while pre-merge it read (unchanged since '94):

/* Accept NTRAPS traps from the inferior.  */

void
startup_inferior (ntraps)
      int ntraps;
{
   int pending_execs = ntraps;
   int terminal_initted;

   /* The process was started by the fork that created it,
      but it will have stopped one instruction after execing the shell.
      Here we must get it up to actual execution of the real program.  */

   clear_proceed_status ();

   init_wait_for_inferior ();

   terminal_initted = 0;

#ifdef STARTUP_INFERIOR
   STARTUP_INFERIOR (pending_execs);
#else
   while (1)
     {
       stop_soon_quietly = 1;    /* Make wait_for_inferior be quiet */
       wait_for_inferior ();
       if (stop_signal != TARGET_SIGNAL_TRAP)
         {
           /* Let shell child handle its own signals in its own way */
           /* FIXME, what if child has exit()ed?  Must exit loop somehow */
           resume (0, stop_signal);
         }
       else
         {
           /* We handle SIGTRAP, however; it means child did an exec.  */
           if (!terminal_initted)
             {
               /* Now that the child has exec'd we know it has already 
set its
                  process group.  On POSIX systems, tcsetpgrp will fail with
                  EPERM if we try it before the child's setpgid.  */

               /* Set up the "saved terminal modes" of the inferior
                  based on what modes we are starting it with.  */
               target_terminal_init ();

               /* Install inferior's terminal modes.  */
               target_terminal_inferior ();

               terminal_initted = 1;
             }
           if (0 == --pending_execs)
             break;
           resume (0, TARGET_SIGNAL_0);          /* Just make it go on */
         }
     }
#endif /* STARTUP_INFERIOR */
   stop_soon_quietly = 0;
}

Re: [PATCH] Use multi-arch'd START_INFERIOR_TRAPS_EXPECTED on Alpha target

[Daniel Jacobowitz]

On Wed, Apr 24, 2002 at 08:06:31PM -0400, Andrew Cagney wrote:
> >The following puts STARTUP_WITH_SHELL
> 
> Hmm, HP merge.  An intermediate version looked like:
> 
>   /* If STARTUP_WITH_SHELL is set, GDB's "run"
> !  * will attempts to start up the debugee under a shell.
> !  * This is in order for argument-expansion to occur. E.g.,
> !  * (gdb) run *
> !  * The "*" gets expanded by the shell into a list of files.
> !  * While this is a nice feature, it turns out to interact badly
> !  * with some of the catch-fork/catch-exec features we have added.
> !  * In particular, if the shell does any fork/exec's before
> !  * the exec of the target program, that can confuse GDB.
> !  * To disable this feature, set STARTUP_WITH_SHELL to 0.
> !  * To enable this feature, set STARTUP_WITH_SHELL to 1.
> !  * The catch-exec traps expected during start-up will
> !  * be 1 if target is not started up with a shell, 2 if it is.
> !  * - RT
> !  */
>   #define STARTUP_WITH_SHELL 1
> - #define START_INFERIOR_TRAPS_EXPECTED (STARTUP_WITH_SHELL + 1)
> 
> Does STARTUP_WITH_SHELL need to be multi-arched?
> 
> I'm wondering if it would be better to make it a variable (``set 
> startup-with-shell <boolean>'').  Looking at its uses it appears that 
> fork-child.c:startup_inferior() would still work (if it did previously).

Silly question.... are there any (supported?  working?) uses for this
besides globbing and backtick interpolation?  I think there aren't, and
I think it would simplify GDB to just have a function which called
glob() and invoked subshells for ``.  It's a little tricky, but not
very.  We'd lose access to things like shell-specific globbing tricks,
but I think that's a worthwhile price to pay.

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

Re: [PATCH] Use multi-arch'd START_INFERIOR_TRAPS_EXPECTED on Alpha target

[Michael Snyder]

Daniel Jacobowitz wrote:
> 
> On Wed, Apr 24, 2002 at 08:06:31PM -0400, Andrew Cagney wrote:
> > >The following puts STARTUP_WITH_SHELL
> >
> > Hmm, HP merge.  An intermediate version looked like:
> >
> >   /* If STARTUP_WITH_SHELL is set, GDB's "run"
> > !  * will attempts to start up the debugee under a shell.
> > !  * This is in order for argument-expansion to occur. E.g.,
> > !  * (gdb) run *
> > !  * The "*" gets expanded by the shell into a list of files.
> > !  * While this is a nice feature, it turns out to interact badly
> > !  * with some of the catch-fork/catch-exec features we have added.
> > !  * In particular, if the shell does any fork/exec's before
> > !  * the exec of the target program, that can confuse GDB.
> > !  * To disable this feature, set STARTUP_WITH_SHELL to 0.
> > !  * To enable this feature, set STARTUP_WITH_SHELL to 1.
> > !  * The catch-exec traps expected during start-up will
> > !  * be 1 if target is not started up with a shell, 2 if it is.
> > !  * - RT
> > !  */
> >   #define STARTUP_WITH_SHELL 1
> > - #define START_INFERIOR_TRAPS_EXPECTED (STARTUP_WITH_SHELL + 1)
> >
> > Does STARTUP_WITH_SHELL need to be multi-arched?
> >
> > I'm wondering if it would be better to make it a variable (``set
> > startup-with-shell <boolean>'').  Looking at its uses it appears that
> > fork-child.c:startup_inferior() would still work (if it did previously).
> 
> Silly question.... are there any (supported?  working?) uses for this
> besides globbing and backtick interpolation?  I think there aren't, and
> I think it would simplify GDB to just have a function which called
> glob() and invoked subshells for ``.  It's a little tricky, but not
> very.  We'd lose access to things like shell-specific globbing tricks,
> but I think that's a worthwhile price to pay.

I'm not so sure.  Sometimes getting the exact behavior of a specific
shell might be important.  Also, might this functionality be used
for piping and I/O redirection?

Re: [PATCH] Use multi-arch'd START_INFERIOR_TRAPS_EXPECTED on Alpha target

[Daniel Jacobowitz]

On Thu, Apr 25, 2002 at 02:17:30PM -0700, Michael Snyder wrote:
> Daniel Jacobowitz wrote:
> > Silly question.... are there any (supported?  working?) uses for this
> > besides globbing and backtick interpolation?  I think there aren't, and
> > I think it would simplify GDB to just have a function which called
> > glob() and invoked subshells for ``.  It's a little tricky, but not
> > very.  We'd lose access to things like shell-specific globbing tricks,
> > but I think that's a worthwhile price to pay.
> 
> I'm not so sure.  Sometimes getting the exact behavior of a specific
> shell might be important.  Also, might this functionality be used
> for piping and I/O redirection?

Oh, I believe you're right.  Probably not a viable idea then.

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