winedbg error

[Dave Hawkes]

Hi all,

I keep getting an assertion in winedbg, before I try and find out the cause
does anyone know the possible reason for this?

winedbg: types.c:573: DEBUG_CopyFieldList: Assertion 'dt->type == dt2->type
&& ((dt->type == DT_STRUCT) || (dt->type == DT_ENUM))' failed.

It looks like its an issue with parsing the debug info, but the windows
application debugs just fine with MS Developer Studio 6.

Thanks,
Dave Hawkes

[eric pouech]

Dave Hawkes wrote:
> Hi all,
>
> I keep getting an assertion in winedbg, before I try and find out the cause
> does anyone know the possible reason for this?
>
> winedbg: types.c:573: DEBUG_CopyFieldList: Assertion 'dt->type == dt2->type
> && ((dt->type == DT_STRUCT) || (dt->type == DT_ENUM))' failed.
>
> It looks like its an issue with parsing the debug info, but the windows
> application debugs just fine with MS Developer Studio 6.

can you tell me when this happens ?
are you trying to run some command, or is it while loading some
symbols...
if the later, which library is currently loaded ?

A+

[Dave Hawkes]

This occurs whenever winedbg attempts to load the symbols from a particular
large application that I have, either due to a fault or if I start windbg
directly on the command line. The debug terminal window will start with the
winedbg starting message appearing and will them immediately terminate with
this error.

Dave Hawkes

----- Original Message -----
From: "eric pouech" 
To: "Dave Hawkes" 
Cc: 
Sent: Tuesday, November 20, 2001 1:20 PM
Subject: Re: winedbg error
> Dave Hawkes wrote:
> > Hi all,
> >
> > I keep getting an assertion in winedbg, before I try and find out the

cause
> > does anyone know the possible reason for this?
> >
> > winedbg: types.c:573: DEBUG_CopyFieldList: Assertion 'dt->type ==

dt2->type
> > && ((dt->type == DT_STRUCT) || (dt->type == DT_ENUM))' failed.
> >
> > It looks like its an issue with parsing the debug info, but the windows
> > application debugs just fine with MS Developer Studio 6.
>
> can you tell me when this happens ?
> are you trying to run some command, or is it while loading some
> symbols...
> if the later, which library is currently loaded ?
> A+
>
> Eric Pouech (http://perso.wanadoo.fr/eric.pouech/)
> "The future will be better tomorrow", Vice President Dan Quayle

[eric pouech]

Dave Hawkes wrote:
> This occurs whenever winedbg attempts to load the symbols from a particular
> large application that I have, either due to a fault or if I start windbg
> directly on the command line. The debug terminal window will start with the
> winedbg starting message appearing and will them immediately terminate with
> this error.

ok, sounds like winedbg is loading MSC debug info and fails...
you could remove the assert in types.c (for DEBUG_CopyFieldlist)
another quick test would be to print, in DEBUG_CopyFieldlist, the
reasons
of the failed assertion:
  if (!(dt->type == dt2->type && ((dt->type == DT_STRUCT) || (dt->type
== DT_ENUM))))
  {
    DEBUG_Printf(DBG_CHN_MESG, "Copyfield list mismatch (%d<>%d): ",
dt->type, dt2->type);
    DEBUG_PrintTypeCast(dt);
    DEBUG_Printf(DBG_CHN_MESG, " ");
    DEBUG_PrintTypeCast(dt2);
    DEBUG_Printf(DBG_CHN_MESG, "\n");
  }
just to see what gives...

[Dave Hawkes]

Hi Eric,

After some experimenting along the lines that you suggested
(DEBUG_PrintTypeCast faulted!), it may be that the problem is something to
do with the 'bool' type as I now eventually get lots of errors like this:

Unknown type '1'
failure on _Bool:t(0,21)=@s8;-16; at 16;

Dave

----- Original Message -----
From: "eric pouech" 
To: "Dave Hawkes" 
Cc: 
Sent: Tuesday, November 20, 2001 3:12 PM
Subject: Re: winedbg error
> Dave Hawkes wrote:
> > This occurs whenever winedbg attempts to load the symbols from a

particular
> > large application that I have, either due to a fault or if I start

windbg
> > directly on the command line. The debug terminal window will start with

the
> > winedbg starting message appearing and will them immediately terminate

with
> > this error.
>
> ok, sounds like winedbg is loading MSC debug info and fails...
> you could remove the assert in types.c (for DEBUG_CopyFieldlist)
> another quick test would be to print, in DEBUG_CopyFieldlist, the
> reasons
> of the failed assertion:
>   if (!(dt->type == dt2->type && ((dt->type == DT_STRUCT) || (dt->type
> == DT_ENUM))))
>   {
> DEBUG_Printf(DBG_CHN_MESG, "Copyfield list mismatch (%d<>%d): ",
> dt->type, dt2->type);
> DEBUG_PrintTypeCast(dt);
> DEBUG_Printf(DBG_CHN_MESG, " ");
> DEBUG_PrintTypeCast(dt2);
> DEBUG_Printf(DBG_CHN_MESG, "\n");
>   }
> just to see what gives...
> --
>
> Eric Pouech (http://perso.wanadoo.fr/eric.pouech/)
> "The future will be better tomorrow", Vice President Dan Quayle

[James Juran]

On Wed, 21 Nov 2001, Dave Hawkes wrote:
> Hi Eric,
>
> After some experimenting along the lines that you suggested
> (DEBUG_PrintTypeCast faulted!), it may be that the problem is something to
> do with the 'bool' type as I now eventually get lots of errors like this:
>
> Unknown type '1'
> failure on _Bool:t(0,21)=@s8;-16; at 16;

Did you compile with GCC-3.X?  I've gotten the same errors when using 
winedbg on a version of Wine compiled with Red Hat's gcc3 package in Red 
Hat 7.2.  I think that gcc-3.X is known to have a different debugging 
format than previous versions.

James

--
James Juran
[e-mail]

[Dave Hawkes]

Yes, wine is built with GCC3...

Dave Hawkes

----- Original Message -----
From: "James Juran" 
To: "Dave Hawkes" 
Cc: "eric pouech" ; 
Sent: Wednesday, November 21, 2001 4:20 PM
Subject: Re: winedbg error
> On Wed, 21 Nov 2001, Dave Hawkes wrote:
> > Hi Eric,
> >
> > After some experimenting along the lines that you suggested
> > (DEBUG_PrintTypeCast faulted!), it may be that the problem is something

to
> > do with the 'bool' type as I now eventually get lots of errors like

this:
> > Unknown type '1'
> > failure on _Bool:t(0,21)=@s8;-16; at 16;
>
> Did you compile with GCC-3.X?  I've gotten the same errors when using
> winedbg on a version of Wine compiled with Red Hat's gcc3 package in Red
> Hat 7.2.  I think that gcc-3.X is known to have a different debugging
> format than previous versions.
>
> James
>
> --
> James Juran
> [e-mail]

[eric pouech]

> After some experimenting along the lines that you suggested
> (DEBUG_PrintTypeCast faulted!),

hmmm may be the type isn't fully initialized... strange indeed
> it may be that the problem is something to
> do with the 'bool' type as I now eventually get lots of errors like this:

would the attached patch fix is someway ?
if it does, I'll provide a more elaborate fix.

A+

[Dave Hawkes]

Yes, this partially fixed one warning message (which could be due to GCC3),
though I still get the type mismatches.

Example type mismatch (from your previous debugging code):

Copyfield list mismatch (4<>1): enum
CDecBorderDialogBarPage::__unnamed --invalid--*

The --invalid-- is from code I added to prevent a NULL pointer being used
(the in memory type info for the pointer was invalid).

I also now only get this warning after the type mismatch warning.

failure on _Bool:t(0,21)=@s8;-16; at ;

Dave Hawkes

----- Original Message -----
From: "eric pouech" 
To: "Dave Hawkes" 
Cc: 
Sent: Wednesday, November 21, 2001 5:03 PM
Subject: Re: winedbg error
> > After some experimenting along the lines that you suggested
> > (DEBUG_PrintTypeCast faulted!),
>
> hmmm may be the type isn't fully initialized... strange indeed
> > it may be that the problem is something to
> > do with the 'bool' type as I now eventually get lots of errors like

this:
> would the attached patch fix is someway ?
> if it does, I'll provide a more elaborate fix.
> A+
>
> Eric Pouech (http://perso.wanadoo.fr/eric.pouech/)
> "The future will be better tomorrow", Vice President Dan Quayle

----------------------------------------------------------------------------
----
> Index: stabs.c
>
> RCS file: /usr/share/cvs/cvsroot/wine/wine/debugger/stabs.c,v
> retrieving revision 1.41
> diff -u -r1.41 stabs.c
> --- stabs.c 2001/08/10 22:29:21 1.41
> +++ stabs.c 2001/11/21 22:01:31
> @@ -516,8 +516,24 @@
>       new_dt = DEBUG_NewDataType(lo, ptd->buf + idx);
>       ptd->idx = idx;
>       break;
> - case '-':
> -
> if (DEBUG_PTS_ReadNum(ptd, &lo) == -1) {
> - DEBUG_Printf(DBG_CHN_MESG, "Should be a number...\n");
> - return -1;
> -
> }
> -
> switch (lo)
> -
> {
> -
> case  1: new_dt = DEBUG_TypeInt; break;
> -
> case  7: new_dt = DEBUG_TypeShortUInt; break;
> -
> case 16: new_dt = DEBUG_TypeUSInt; break;
> -
> case 19: new_dt = DEBUG_TypeString; break;
> -
> default:
> - DEBUG_Printf(DBG_CHN_MESG, "Unsupported integral type (%d)\n", lo);
> - return -1;
> -
> }
> -
> break;
> default:
> -
> DEBUG_Printf(DBG_CHN_MESG, "Unknown type '%c'\n", *ptd->ptr);
> -
> DEBUG_Printf(DBG_CHN_MESG, "Unknown type '%c'\n", ptd->ptr[-1]);
> return -1;
> }
>    }

[eric pouech]

Dave Hawkes wrote:
> Yes, this partially fixed one warning message (which could be due to GCC3),
> though I still get the type mismatches.

well, the warning you got (in Bool type) may be caused by GCC3 (some
other
stabs information), but not the one in Copyfield. this one comes from an
error in MSC debug info, so GCC shouldn't play role (except if there's
memory trashing)
> Example type mismatch (from your previous debugging code):
>
> Copyfield list mismatch (4<>1): enum
> :__unnamed --invalid--*

hmm this sounds like C++ code... which winedbg doesn't currently handle
properly... there's also lots of unsupported features in MSC debug info
loading (see debugger/msc.c for an outlook)
(could you anyway tell me the def of CDecBorderDialogBarPage)

this will be rather hard to debug and fix remotely...
> I also now only get this warning after the type mismatch warning.
>
> failure on _Bool:t(0,21)=@s8;-16; at ;

well, it seems we need to chew the ';' too (but this won't fix the issue
above)
add 	    
    if (*ptd->ptr++ != ';') return -1;
in the patch I already sent you (here's the whole modif)

case '-':
        if (DEBUG_PTS_ReadNum(ptd, &lo) == -1) {
    	DEBUG_Printf(DBG_CHN_MESG, "Should be a number...\n");
    	return -1;
        }
        switch (lo)
        {
        case  1: new_dt = DEBUG_TypeInt;		break;
        case  7: new_dt = DEBUG_TypeShortUInt;	break;
        case 16: new_dt = DEBUG_TypeUSInt;		break;
        case 19: new_dt = DEBUG_TypeString;		break;
        default:
    	DEBUG_Printf(DBG_CHN_MESG, "Unsupported integral type (%d)\n", lo);
    	return -1;
        }
        if (*ptd->ptr++ != ';') return -1;
        break;

[Dave Hawkes]

This now fixes all the warnings due to GCC3, so this patch could now be
submitted.

I still have the problem with the main application C++ code, everytime I
rebuild the application it gives a warning on a different class or
structure, so the problem may be related to the size of the application (It
takes an hour to build on 2 x 1200Mhz machine).

Now that I have got the debugger to ignore these errors it continues load
the application, but eventually freezes without any warning or error at some
point during the rest of the loading process. I'll try and break into gdb to
find out what else gives now...

Thanks
Dave Hawkes

----- Original Message -----
From: "eric pouech" 
To: "Dave Hawkes" 
Cc: 
Sent: Thursday, November 22, 2001 1:04 PM
Subject: Re: winedbg error
> Dave Hawkes wrote:
> > Yes, this partially fixed one warning message (which could be due to

GCC3),
> > though I still get the type mismatches.
>
> well, the warning you got (in Bool type) may be caused by GCC3 (some
> other
> stabs information), but not the one in Copyfield. this one comes from an
> error in MSC debug info, so GCC shouldn't play role (except if there's
> memory trashing)
> > Example type mismatch (from your previous debugging code):
> >
> > Copyfield list mismatch (4<>1): enum
> > :__unnamed --invalid--*
>
> hmm this sounds like C++ code... which winedbg doesn't currently handle
> properly... there's also lots of unsupported features in MSC debug info
> loading (see debugger/msc.c for an outlook)
> (could you anyway tell me the def of CDecBorderDialogBarPage)
>
> this will be rather hard to debug and fix remotely...
> > I also now only get this warning after the type mismatch warning.
> >
> > failure on _Bool:t(0,21)=@s8;-16; at ;
>
> well, it seems we need to chew the ';' too (but this won't fix the issue
> above)
> add
> if (*ptd->ptr++ != ';') return -1;
> in the patch I already sent you (here's the whole modif)
>
> case '-':
>     if (DEBUG_PTS_ReadNum(ptd, &lo) == -1) {
> DEBUG_Printf(DBG_CHN_MESG, "Should be a number...\n");
> return -1;
>     }
>     switch (lo)
>     {
>     case  1: new_dt = DEBUG_TypeInt; break;
>     case  7: new_dt = DEBUG_TypeShortUInt; break;
>     case 16: new_dt = DEBUG_TypeUSInt; break;
>     case 19: new_dt = DEBUG_TypeString; break;
>     default:
> DEBUG_Printf(DBG_CHN_MESG, "Unsupported integral type (%d)\n", lo);
> return -1;
>     }
>     if (*ptd->ptr++ != ';') return -1;
>     break;
> --
>
> Eric Pouech (http://perso.wanadoo.fr/eric.pouech/)
> "The future will be better tomorrow", Vice President Dan Quayle