I sought more deeply and after analysing the map file of the linker ,
I find that the break of variable occur when I get message from my IPC
system V message queue !
The msgrcv function can write one byte ( perhaps more ??) after the
buffer , and only with the arm-gcc !
See under : my original get message function from the message queue :
//---------------------------------------------------------------------------//
// Extract message in message queue
//---------------------------------------------------------------------------//
TBOOL bFFMSG_ReceiveMsg( T_FMSG_OBJECT * ptObjFm, key_t tKeyFm , long
lType )
{
// Get ID of mq
int iFileMsgID = msgget( tKeyFm, 0700 );
// Get Msg in mq
if( msgrcv( iFileMsgID, &ptObjFm , sizeof(T_FMSG_OBJECT), lType,
IPC_NOWAIT ) < 0) return (aFALSE);
// succes
return (aTRUE);
}
See under : the test get message function from the message queue :
//---------------------------------------------------------------------------//
// Extract message in message queue ( with local protection and test
program)
/ / => result :
// ### msgrcv test program ###
// ## 00 ## 53 ## 53 ## 53 ## 53
//---------------------------------------------------------------------------//
TBOOL bFFMSG_ReceiveMsg( T_FMSG_OBJECT * ptObjFm, key_t tKeyFm , long
lType )
{
T_FMSG_OBJECT tObjFm;
TUCHAR tcProtect[] = {0x53, 0x53, 0x53, 0x53, 0x53 }
// Get ID of mq
int iFileMsgID = msgget( tKeyFm, 0700 );
printf("### msgrcv test program ###\r\n");
// Get Msg in mq
if( msgrcv( iFileMsgID, &tTmp.tObjFm , sizeof(T_FMSG_OBJECT), lType,
IPC_NOWAIT ) < 0) return (aFALSE);
printf("## %02x ## %02x ## %02x ## %02x ## %02x \r\n, tcProtect[0],
tcProtect[1], tcProtect[2], tcProtect[3], tcProtect[4], )
// Copy Msg to user
memcpy( ptObjFm, &tTmp.tObjFm, sizeof(T_FMSG_OBJECT) );
// succes
return (aTRUE);
}
Nicolas TOUCHARD a écrit :
> Marc Kleine-Budde a écrit :
>
>> Nicolas TOUCHARD wrote:
>>
>>
>>> on my host PC gcc 3.3.5
>>> the cross compiler with the ptxdist 0.10.4 : gcc 4.0.2
>>>
>>> other remarks. I'v 2 process and the procces communicate with message
>>> file (IPC sytem V). The bug occur in procces 1 when the other send a
>>> command.
>>>
>>>
>> Have you verified that this work on the PC. Are you sure you don't have
>> a bug in you application.
>>
>> Is there a chance for a (minial) testcase?
>>
>> Marc
>>
>>
>>
>
> Yes I 've verified ( I've jsute repeat the test) and it's work on my
> host PC . I hope it's just a bug :-) , I try to isolate the moment when
> the variable is broke ...
>
>
>
>
--
Nicolas Touchard
NKE electronics
Rue Gutenberg Z.I. de Kerandré
56700 HENNEBONT - FRANCE
Email : [EMAIL PROTECTED]
Tél. : (33) 2-97-36-10-12
Fax : (33) 2-97-36-55-17
--
ptxdist mailing list
[email protected]
