Jarmo Paavilainen wrote:
>
> Hi,
>
> Wrong forum but Im a newbie in Linux and this is a Linux compiler problem
> (works on Watcom C++ 11/ MS VC++ 6.0). If someone knows a better place let
> me know.
linux-gcc is the correct list for such problems.
I can't see anything wrong in this code, but more info may help. Is
'pek' destroyed at the first iteration, or is it destroyed after the
first ? Also, this code assumes that tList.Append makes a new copy of
the object it is appending, is it so?
>
> The code looks something like this:
>
> ///////////////////////////////////////
> //
> const char *pek;
> for( loopIP = IPList.GetFirstNode(); loopIP; loopIP = loopIP->GetNext())
> {
> if( bMajor ) pek = strrchr( loopIP->GetHolder().strDNS, '.' );
> else pek = strchr( loopIP->GetHolder().strDNS, '.' );
> if( !pek ) pek = " <N/A> ";
>
> fprintf( stderr, "1) pek = \"%s\"\n", pek ); // ** Still ok here **
>
> for( loopDNS = tList.GetFirstNode(); loopDNS; loopDNS =
> loopDNS->GetNext())
> {
> if( strcmp( pek, loopDNS->GetHolder().strDNS ) == 0 ) break;
> else fprintf( stderr, "2) pek = \"%s\"\n", pek ); // ** NOT
>here **
> }
>
> fprintf( stderr, "3) pek = \"%s\"\n", pek ); // If point 2 was executed then
> pek is fucked up here
>
> if( loopDNS ) loopDNS->GetPointer()->iCount +=
>loopIP->GetHolder().iCount;
> else
> {
> sDNS sTemp;
> if( strlen( pek ) < sizeof( sTemp.strDNS ) - 1)
> {
> fprintf( stderr, "4) pek = \"%s\"\n", pek ); // If point 2 was executed then
> pek is still bad
> strcpy( sTemp.strDNS, pek );
> sTemp.iCount = loopIP->GetHolder().iCount;
> tList.Append( sTemp );
> fprintf( stderr, "5) pek = \"%s\"\n", pek ); // Pek is the same here as it
> was at point 1, but the bad data is allready copied to sTemp.strDNS
> }
> else fprintf( stderr, "** ERROR ** Overflow detected." );
> }
> }
>
> //
> ///////////////////////////////////////
>
> It seems that "pek" is moved by strcmp(...) to point at bogus data if the
> strings are not the same. But after a few commands the compiler restores the
> value. Maybe strcmp(...) does not restore all registers at exit. And after a
> while the compiler desides to restore them instead.
>
> Or Im I doing something wrong?
>
> // Jarmo
