Re: Mingw GDB build fails for M16C target

Re: Mingw GDB build fails for M16C target

[Kevin Buettner]

[Note: Reply-to has been set to gdb-patches.]

On Sat, 25 Apr 2009 11:05:25 +0530
Mahesh Basavaraj Shavari <Mahesh.Shavari@kpitcummins.com> wrote:

> /home/mahesh/GDB/gdb-6.8/gdb-6.8.50.20080811/sim/m32c/gdb-if.c:541:
> error: `SIGTRAP' undeclared (first use in this function)

Mahesh,

Please try the patch below and let me know if it works for you.

Jim,

It appears to me that comment, "Oddly, the gdb/sim interface uses host
signal numbers...", isn't true.  My patch deletes this comment in
addition to using TARGET_SIGNAL_foo instead of SIGfoo.  If you look at
remote-sim.c, you'll see that `enum target_signal' (which defines the
various TARGET_SIGNAL_foo constants) is used throughout the file.

Okay to apply?

Kevin

	* m32c/gdb-if.c (m32c_signal_to_host): Rename to
	m32c_signal_to_target.  Change signal return values from SIGILL,
	SIGTRAP, SIGSEGV, etc. to TARGET_SIGNAL_ILL, TARGET_SIGNAL_TRAP,
	TARGET_SIGNAL_SEGV, etc.  Fix all callers.

Index: m32c/gdb-if.c
===================================================================
RCS file: /cvs/src/src/sim/m32c/gdb-if.c,v
retrieving revision 1.7
diff -u -p -r1.7 gdb-if.c
--- m32c/gdb-if.c	14 Jan 2009 10:53:07 -0000	1.7
+++ m32c/gdb-if.c	28 Apr 2009 22:38:44 -0000
@@ -523,52 +523,35 @@ int siggnal;
 
 
 /* Given a signal number used by the M32C bsp (that is, newlib),
-   return a host signal number.  (Oddly, the gdb/sim interface uses
-   host signal numbers...)  */
+   return a target signal number used by GDB.  */
 int
-m32c_signal_to_host (int m32c)
+m32c_signal_to_target (int m32c)
 {
   switch (m32c)
     {
     case 4:
-#ifdef SIGILL
-      return SIGILL;
-#else
-      return SIGSEGV;
-#endif
+      return TARGET_SIGNAL_ILL;
 
     case 5:
-      return SIGTRAP;
+      return TARGET_SIGNAL_TRAP;
 
     case 10:
-#ifdef SIGBUS
-      return SIGBUS;
-#else
-      return SIGSEGV;
-#endif
+      return TARGET_SIGNAL_BUS;
 
     case 11:
-      return SIGSEGV;
+      return TARGET_SIGNAL_SEGV;
 
     case 24:
-#ifdef SIGXCPU
-      return SIGXCPU;
-#else
-      break;
-#endif
+      return TARGET_SIGNAL_XCPU;
 
     case 2:
-      return SIGINT;
+      return TARGET_SIGNAL_INT;
 
     case 8:
-#ifdef SIGFPE
-      return SIGFPE;
-#else
-      break;
-#endif
+      return TARGET_SIGNAL_FPE;
 
     case 6:
-      return SIGABRT;
+      return TARGET_SIGNAL_ABRT;
     }
 
   return 0;
@@ -588,7 +571,7 @@ handle_step (int rc)
   else if (M32C_STOPPED (rc))
     {
       reason = sim_stopped;
-      siggnal = m32c_signal_to_host (M32C_STOP_SIG (rc));
+      siggnal = m32c_signal_to_target (M32C_STOP_SIG (rc));
     }
   else
     {

Re: Mingw GDB build fails for M16C target

[Daniel Jacobowitz]

On Tue, Apr 28, 2009 at 03:57:24PM -0700, Kevin Buettner wrote:
> Jim,
> 
> It appears to me that comment, "Oddly, the gdb/sim interface uses host
> signal numbers...", isn't true.  My patch deletes this comment in
> addition to using TARGET_SIGNAL_foo instead of SIGfoo.  If you look at
> remote-sim.c, you'll see that `enum target_signal' (which defines the
> various TARGET_SIGNAL_foo constants) is used throughout the file.

Shouldn't this be the same as common/sim-signal.c?  sim_signal_to_host
uses native numbers and sim_signal_to_target uses GDB protocol
numbers.

Then again, I'm totally baffled by this bit of the sim.  It looks like
sim_stop_reason returns a target signal and the callers check it
against a host signal???

-- 
Daniel Jacobowitz
CodeSourcery

Re: Mingw GDB build fails for M16C target

[Kevin Buettner]

On Tue, 28 Apr 2009 22:35:11 -0400
Daniel Jacobowitz <drow@false.org> wrote:

> On Tue, Apr 28, 2009 at 03:57:24PM -0700, Kevin Buettner wrote:
> > Jim,
> > 
> > It appears to me that comment, "Oddly, the gdb/sim interface uses host
> > signal numbers...", isn't true.  My patch deletes this comment in
> > addition to using TARGET_SIGNAL_foo instead of SIGfoo.  If you look at
> > remote-sim.c, you'll see that `enum target_signal' (which defines the
> > various TARGET_SIGNAL_foo constants) is used throughout the file.
> 
> Shouldn't this be the same as common/sim-signal.c?  sim_signal_to_host
> uses native numbers and sim_signal_to_target uses GDB protocol
> numbers.

It's not clear to me what you're asking about.

If you're asking why m32c/gdb-if.c doesn't use common/sim-signal.c,
that's because the m32c sim doesn't use the infrastructure in common/
at all.

If you're asking why I don't simply copy sim_signal_to_target for
use in m32c/gdb-if.c, that's because the m32c sim doesn't use the
SIM_SIGfoo constants that common/sim-signal.c uses.  It does bother me
though that m32c_signal_to_target() uses hardcoded constants like 4,
5, 10, 11, etc.  I suspect it was done this way because including the
correct newlib header file (without conflict with certain system
headers) was difficult.  That's just a guess though.

> Then again, I'm totally baffled by this bit of the sim.  It looks like
> sim_stop_reason returns a target signal and the callers check it
> against a host signal???

I looked at that yesterday prior to posting my patch.  I thought I
understood what was going on, but now that I look at it closer, I
too am baffled.  I noticed yesterday that signal() is being called to
handle SIGINT.  My thought at the time was that the handler was
somehow manipulating the sim's state so that it would return the sim
specific code for the signal.  But that's not what's happening.  The
handler simply checks to make sure that it (the sim) is not stopped;
if not, it prints a message, and then exits.  So it seems to me that
sim_signal_to_host() and that check that you mention are completely
unnecessary.

Kevin

Re: Mingw GDB build fails for M16C target

[Daniel Jacobowitz]

On Wed, Apr 29, 2009 at 12:50:49PM -0700, Kevin Buettner wrote:
> > Shouldn't this be the same as common/sim-signal.c?  sim_signal_to_host
> > uses native numbers and sim_signal_to_target uses GDB protocol
> > numbers.
> 
> It's not clear to me what you're asking about.
> 
> If you're asking why m32c/gdb-if.c doesn't use common/sim-signal.c,
> that's because the m32c sim doesn't use the infrastructure in common/
> at all.
> 
> If you're asking why I don't simply copy sim_signal_to_target for
> use in m32c/gdb-if.c, that's because the m32c sim doesn't use the
> SIM_SIGfoo constants that common/sim-signal.c uses.  It does bother me
> though that m32c_signal_to_target() uses hardcoded constants like 4,
> 5, 10, 11, etc.  I suspect it was done this way because including the
> correct newlib header file (without conflict with certain system
> headers) was difficult.  That's just a guess though.

I was actually asking about sim_signal_to_host, since this was
m32c_signal_to_host.  The sim_ version uses SIGILL, SIGTRAP, etc.

> > Then again, I'm totally baffled by this bit of the sim.  It looks like
> > sim_stop_reason returns a target signal and the callers check it
> > against a host signal???
> 
> I looked at that yesterday prior to posting my patch.  I thought I
> understood what was going on, but now that I look at it closer, I
> too am baffled.  I noticed yesterday that signal() is being called to
> handle SIGINT.  My thought at the time was that the handler was
> somehow manipulating the sim's state so that it would return the sim
> specific code for the signal.  But that's not what's happening.  The
> handler simply checks to make sure that it (the sim) is not stopped;
> if not, it prints a message, and then exits.  So it seems to me that
> sim_signal_to_host() and that check that you mention are completely
> unnecessary.

Yes... I don't think any of the conversion routines that use host
numbers should be there.  I'm not feeling ambitious enough to remove
them from all sims, but perhaps you can get just m32c?

-- 
Daniel Jacobowitz
CodeSourcery

Re: Mingw GDB build fails for M16C target

[Kevin Buettner]

On Wed, 29 Apr 2009 16:01:24 -0400
Daniel Jacobowitz <drow@false.org> wrote:

> > If you're asking why I don't simply copy sim_signal_to_target for
> > use in m32c/gdb-if.c, that's because the m32c sim doesn't use the
> > SIM_SIGfoo constants that common/sim-signal.c uses.  It does bother me
> > though that m32c_signal_to_target() uses hardcoded constants like 4,
> > 5, 10, 11, etc.  I suspect it was done this way because including the
> > correct newlib header file (without conflict with certain system
> > headers) was difficult.  That's just a guess though.
> 
> I was actually asking about sim_signal_to_host, since this was
> m32c_signal_to_host.  The sim_ version uses SIGILL, SIGTRAP, etc.

Wow, did I guess wrong.

My patch changes the name from m32c_signal_to_host() to
m32c_signal_to_target() and switches from use of host-based numbers
to target-specific numbers.

> Yes... I don't think any of the conversion routines that use host
> numbers should be there.  I'm not feeling ambitious enough to remove
> them from all sims, but perhaps you can get just m32c?

Yes, I have done this.  See above.

Kevin

Re: Mingw GDB build fails for M16C target

[Daniel Jacobowitz]

On Wed, Apr 29, 2009 at 01:27:02PM -0700, Kevin Buettner wrote:
> > Yes... I don't think any of the conversion routines that use host
> > numbers should be there.  I'm not feeling ambitious enough to remove
> > them from all sims, but perhaps you can get just m32c?
> 
> Yes, I have done this.  See above.

Gotcha now!  Sounds good to me.

-- 
Daniel Jacobowitz
CodeSourcery

Re: Mingw GDB build fails for M16C target

[Kevin Buettner]

On Tue, 28 Apr 2009 15:57:24 -0700
Kevin Buettner <kevinb@redhat.com> wrote:

> 	* m32c/gdb-if.c (m32c_signal_to_host): Rename to
> 	m32c_signal_to_target.  Change signal return values from SIGILL,
> 	SIGTRAP, SIGSEGV, etc. to TARGET_SIGNAL_ILL, TARGET_SIGNAL_TRAP,
> 	TARGET_SIGNAL_SEGV, etc.  Fix all callers.

Jim sent me an approval for this patch.

Committed.

Kevin