[commit] Fix flags type initialization bug

[commit] Fix flags type initialization bug

[Mark Kettenis]

I'm working on printing proper flags for some registers on SPARC when
I hit this.  It makes gdb crash, so perhaps I should put this on the
release branch too?

Mark


Index: ChangeLog
from  Mark Kettenis  <kettenis@gnu.org>

	* gdbtypes.c (init_flags_type): Set all fields to zero instead of
	just the first one.

Index: gdbtypes.c
===================================================================
RCS file: /cvs/src/src/gdb/gdbtypes.c,v
retrieving revision 1.105
diff -u -p -r1.105 gdbtypes.c
--- gdbtypes.c 1 Mar 2006 19:34:46 -0000 1.105
+++ gdbtypes.c 22 Aug 2006 19:41:50 -0000
@@ -878,7 +878,7 @@ init_flags_type (char *name, int length)
   type = init_type (TYPE_CODE_FLAGS, length, TYPE_FLAG_UNSIGNED, name, NULL);
   TYPE_NFIELDS (type) = nfields;
   TYPE_FIELDS (type) = TYPE_ALLOC (type, nfields * sizeof (struct field));
-  memset (TYPE_FIELDS (type), 0, sizeof (struct field));
+  memset (TYPE_FIELDS (type), 0, nfields * sizeof (struct field));
 
   return type;
 }

Re: [commit] Fix flags type initialization bug

[Joel Brobecker]

> I'm working on printing proper flags for some registers on SPARC when
> I hit this.  It makes gdb crash, so perhaps I should put this on the
> release branch too?

Is there a way to cause GDB 6.5 to crash, or is only after you
applied your changes?

Right now, I don't have any plan in making a minor release from
the 6.5 branch, but I don't mind making one if the fixes of this
kind keep piling up.


> Index: ChangeLog
> from  Mark Kettenis  <kettenis@gnu.org>
> 
> 	* gdbtypes.c (init_flags_type): Set all fields to zero instead of
> 	just the first one.
> 
> Index: gdbtypes.c
> ===================================================================
> RCS file: /cvs/src/src/gdb/gdbtypes.c,v
> retrieving revision 1.105
> diff -u -p -r1.105 gdbtypes.c
> --- gdbtypes.c 1 Mar 2006 19:34:46 -0000 1.105
> +++ gdbtypes.c 22 Aug 2006 19:41:50 -0000
> @@ -878,7 +878,7 @@ init_flags_type (char *name, int length)
>    type = init_type (TYPE_CODE_FLAGS, length, TYPE_FLAG_UNSIGNED, name, NULL);
>    TYPE_NFIELDS (type) = nfields;
>    TYPE_FIELDS (type) = TYPE_ALLOC (type, nfields * sizeof (struct field));
> -  memset (TYPE_FIELDS (type), 0, sizeof (struct field));
> +  memset (TYPE_FIELDS (type), 0, nfields * sizeof (struct field));
>  
>    return type;
>  }

-- 
Joel

Re: [commit] Fix flags type initialization bug

[Mark Kettenis]

> Date: Tue, 22 Aug 2006 13:41:19 -0700
> From: Joel Brobecker <brobecker@adacore.com>
> 
> > I'm working on printing proper flags for some registers on SPARC when
> > I hit this.  It makes gdb crash, so perhaps I should put this on the
> > release branch too?
> 
> Is there a way to cause GDB 6.5 to crash, or is only after you
> applied your changes?

This is used in i386-tdep.c since january or so, and I think there is
a risk of having it crash there, but I've never seen it happen.

> Right now, I don't have any plan in making a minor release from
> the 6.5 branch, but I don't mind making one if the fixes of this
> kind keep piling up.

This one alone isn't enough, but if there are a few others like it
that can potentially crash gdb itself, I think we should consider
having a 6.5.1 release.

Mark

Re: [commit] Fix flags type initialization bug

[Joel Brobecker]

> > Is there a way to cause GDB 6.5 to crash, or is only after you
> > applied your changes?
> 
> This is used in i386-tdep.c since january or so, and I think there is
> a risk of having it crash there, but I've never seen it happen.

If there is a risk, then let's apply the patch. Let me know if you'd
like me to do it. I have a check out of the branch already available,
so I can do it pretty efficiently.

> > Right now, I don't have any plan in making a minor release from
> > the 6.5 branch, but I don't mind making one if the fixes of this
> > kind keep piling up.
> 
> This one alone isn't enough, but if there are a few others like it
> that can potentially crash gdb itself, I think we should consider
> having a 6.5.1 release.

You're no3 on the list of fixes that made it to the branch.
I try to be careful in my review of the patches that are applied,
so hopefully I didn't miss any. My idea was to make a 6.5.1 release
in october or november if I had enough patches. 3 could qualify as
enough. Maybe we'll have a couple more by then.

-- 
Joel

Re: [commit] Fix flags type initialization bug

[Mark Kettenis]

> Date: Tue, 22 Aug 2006 22:11:22 -0700
> From: Joel Brobecker <brobecker@adacore.com>
> 
> > > Is there a way to cause GDB 6.5 to crash, or is only after you
> > > applied your changes?
> > 
> > This is used in i386-tdep.c since january or so, and I think there is
> > a risk of having it crash there, but I've never seen it happen.
> 
> If there is a risk, then let's apply the patch. Let me know if you'd
> like me to do it. I have a check out of the branch already available,
> so I can do it pretty efficiently.

That'd be great; I don't have such a tree and I'm a bit short of
diskspace right now.

Re: [commit] Fix flags type initialization bug

[Joel Brobecker]

Hi Mark,

> > If there is a risk, then let's apply the patch. Let me know if you'd
> > like me to do it. I have a check out of the branch already available,
> > so I can do it pretty efficiently.
> 
> That'd be great; I don't have such a tree and I'm a bit short of
> diskspace right now.

Sorry for the delay, I was away traveling for a week. I just applied
your patch to the branch:

from  Mark Kettenis  <kettenis@gnu.org>

        * gdbtypes.c (init_flags_type): Set all fields to zero instead of
        just the first one.

Cheers,
-- 
Joel

Re: [commit] Fix flags type initialization bug

[Mark Kettenis]

> Date: Fri, 1 Sep 2006 12:01:55 -0700
> From: Joel Brobecker <brobecker@adacore.com>
> 
> Hi Mark,
> 
> > > If there is a risk, then let's apply the patch. Let me know if you'd
> > > like me to do it. I have a check out of the branch already available,
> > > so I can do it pretty efficiently.
> > 
> > That'd be great; I don't have such a tree and I'm a bit short of
> > diskspace right now.
> 
> Sorry for the delay, I was away traveling for a week. I just applied
> your patch to the branch:
> 
> from  Mark Kettenis  <kettenis@gnu.org>
> 
>         * gdbtypes.c (init_flags_type): Set all fields to zero instead of
>         just the first one.

Thanks!

Mark