This loop never ends in some situation, Ron can you look at it?


/* Step 3 - Call Cleanup Functions */

pAlloc = s_pCurrBlock;
do
{
if( s_pCurrBlock->used == s_uUsedFlag )
{
s_pCurrBlock->used |= HB_GC_DELETE;

/* call the cleanup function - now for NON Blosks. */
if( s_pCurrBlock->pFunc )
{
HB_TRACE( HB_TR_INFO, ( "Cleanup, %p", s_pCurrBlock ) );
( s_pCurrBlock->pFunc )( ( void * ) ( s_pCurrBlock + 1 ) );
HB_TRACE( HB_TR_INFO, ( "DONE Cleanup, %p", s_pCurrBlock ) );
}
}

s_pCurrBlock = s_pCurrBlock->pNext;
}
while( s_pCurrBlock && ( s_pCurrBlock != pAlloc ) );

--
Enrico Maria Giordano

http://www.emagsoftware.it
http://www.emagsoftware.it/emgmusic
http://www.emagsoftware.it/spectrum
http://www.emagsoftware.it/tbosg

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Il 22/08/2024 17:53, Enrico Maria Giordano ha scritto:

This loop never ends in some situation, Ron can you look at it?

Please note that I can reproduce the problem (with a FWH/MSC64 sample),
so any suggestions is very welcome. This is the sample:

#include "Fivewin.ch"


FUNCTION MAIN()

LOCAL oServer := MARIA_CONNECT( "access data", .T. )

TEST( oServer )

? "Before HB_GCALL()"

HB_GCALL( .F. )

? "After HG_GCALL()"

oServer:Close()

? "ALL OK"

RETURN NIL

STATIC FUNCTION TEST( oServer )

LOCAL oRs := oServer:RowSet( "Select 1" )

oRs:End()

RETURN NIL

With a call to HB_GCALL( .F. ) after oRs:End() the problem goes away.

--
Enrico Maria Giordano

http://www.emagsoftware.it
http://www.emagsoftware.it/emgmusic
http://www.emagsoftware.it/spectrum
http://www.emagsoftware.it/tbosg

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Il 22/08/2024 17:53, Enrico Maria Giordano ha scritto:

            ( s_pCurrBlock->pFunc )( ( void * ) ( s_pCurrBlock + 1 ) );

Commenting out this line the problem goes away. Just to give you as many
hints as possible.

--
Enrico Maria Giordano

http://www.emagsoftware.it
http://www.emagsoftware.it/emgmusic
http://www.emagsoftware.it/spectrum
http://www.emagsoftware.it/tbosg

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

This is from garbage.c of Harbour, could it be the problem?

/*
* infinite loop can appear when we are executing clean-up functions
* scanning s_pCurrBlock. It's possible that one of them will free
* the GC block which we are using as stop condition. Only blocks
* for which we set HB_GC_DELETE flag are guarded against releasing.
* To avoid such situation first we are moving blocks which will be
* deleted to separate list. It's additional operation but it can
* even increase the speed when we are deleting only few percent
* of all allocated blocks because in next passes we will scan only
* deleted block list. [druzus]
*/

--
Enrico Maria Giordano

http://www.emagsoftware.it
http://www.emagsoftware.it/emgmusic
http://www.emagsoftware.it/spectrum
http://www.emagsoftware.it/tbosg

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

The crash is here:

hb_vmSend( 0 );

It is inside classes.c, function hb_clsFinalize(), any ideas?

void hb_clsFinalize( PHB_ITEM pObject )
{
HB_THREAD_STUB

USHORT uiClass;

if( HB_IS_OBJECT( pObject ) )
uiClass = pObject->item.asArray.value->uiClass;
else
/* TODO Error!
*/
return;

if( uiClass && uiClass <= s_uiClasses )
{
PCLASS pClass = s_pClasses + ( uiClass - 1 );

if( pClass->pDestructor )
{
if( pClass->uiScope & HB_OO_CLS_DESTRUC_SYMB )
{
assert( hb_stack_ready );

if( s_AllowDestructors /* && hb_stack_ready */ )
{
if( HB_IS_SYMBOL( hb_stackBaseItem() ) &&
(
hb_stackBaseItem()->item.asSymbol.pCargo->uiSuperClass == 0 || ( USHORT
) hb_stackBaseItem()->item.asSymbol.pCargo->uiSuperClass != uiClass ||
( HB_IS_ARRAY( hb_stackSelfItem() ) && hb_stackSelfItem()->item.asArray.value != pObject->item.asArray.value )
) && strcmp( hb_stackBaseItem()->item.asSymbol.value->szName,
"__CLSINSTSUPER" ) )
{
/* To DISABLE GC here where no refernce to this
object will cause GPF for double release!
*/
BOOL bCollecting = hb_gcSetCollecting( TRUE ),
bPop = TRUE;
PHB_SYMB pDestructor = pClass->pDestructor->pMessage->pSymbol;

/* Save the existing Return Value and Top Item if any.
*/
if( HB_IS_ARRAY( &HB_VM_STACK.Return ) && HB_VM_STACK.Return.item.asArray.value == pObject->item.asArray.value )
{
/* Don't process HB_VM_STACK.Return!
*/
bPop = FALSE;

/* Save top item which can be processed at this moment
*/
hb_stackPush();
}
else
hb_vmPushState();

hb_vmPushSymbol( pDestructor );
hb_vmPush( pObject ); /* Do NOT Forward!!! */
hb_vmSend( 0 );

/* Restore the existing Return Value and Top Item if any.
*/
if( bPop )
hb_vmPopState();
else
/* Restore top item
*/
hb_stackDec();

hb_gcSetCollecting( bCollecting );
}
}
else
hb_errInternal( HB_EI_ERRUNRECOV, "Destructors disabled! Destructor of class: '%s' can't be executed.", pClass->szName, NULL );
}
}
}
}

--
Enrico Maria Giordano

http://www.emagsoftware.it
http://www.emagsoftware.it/emgmusic
http://www.emagsoftware.it/spectrum
http://www.emagsoftware.it/tbosg

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

pDestructor and pObject contain numbers, so appear valid (how can I
check them?). pDestructor -> szName contains __DESTRUCTOR. What else can
I check?

--
Enrico Maria Giordano

http://www.emagsoftware.it
http://www.emagsoftware.it/emgmusic
http://www.emagsoftware.it/spectrum
http://www.emagsoftware.it/tbosg

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)