Just to make hay while the sun shines: target 3.2

Btw. translate the German comment...

Fixed in CWS mba32issues02

.

The code obviously works fine. Please have a look on the code and verify that
the change is correct.

@mba:  The modification of idl/source/objects/bastype.cxx in
<http://hg.services.openoffice.org/hg/cws/mba32issues02/rev/53b0444724da> is
still not correct:
- it fails to multiply the final result assigned to *pValue by nSign;
- it arbitrarily fails for LONG_MIN;
- technically speaking, it incurs undefined behavior when casting between long
and unsigned long;
- it violates our coding standards (tabs, C-style casts).
As the only call to ReadRangeSvIdl is with nMin == 0, why not simply drop
support for reading negative numbers in the first place and simplify accordingly?

.

Good idea. :-)

So something like that:

static BOOL ReadRangeSvIdl( SvStringHashEntry * pName, SvTokenStream & rInStm,
                            ULONG nMin, ULONG nMax, ULONG* pValue )
{
    UINT32 nTokPos = rInStm.Tell();
    SvToken * pTok = rInStm.GetToken_Next();

    if( pTok->Is( pName ) )
    {
        BOOL bOk = TRUE;
        BOOL bBraket = rInStm.Read( '(' );
        if( bBraket || rInStm.Read( '=' ) )
        {
            short nSign = 1;
            pTok = rInStm.GetToken_Next();
            if( pTok->IsChar() && pTok->GetChar() == '-' )
            {
                nSign = -1;
                pTok = rInStm.GetToken_Next();
            }
            else if( pTok->IsChar() && pTok->GetChar() == '+' )
                pTok = rInStm.GetToken_Next();

            if( pTok->IsInteger() )
            {
                if ( nSign == -1 )
                    return FALSE;

                ULONG n = pTok->GetNumber();
                if ( n >= nMin && n =< nMax )
                {
                    *pValue = n;
                }
                else
                    bOk = FALSE;

                if( bOk && bBraket )
                    bOk = rInStm.Read( ')' );
            }
            else
                bOk = pTok->IsChar() && pTok->GetChar() == ')';
        }
        if( bOk )
            return TRUE;
    }
    rInStm.Seek( nTokPos );
    return FALSE;
}

@mba: looks better (it fails for input "-0" for which the original would have
succeeded, however, and it might make sense to drop scanning for '-' and having
nSign in the code completely)

OK, makes sense. Then it also makes sense not to look for brackets. Looks very
simple now and of course works. Please verify:

static BOOL ReadRangeSvIdl( SvStringHashEntry * pName, SvTokenStream & rInStm,
                            ULONG nMin, ULONG nMax, ULONG* pValue )
{
    UINT32 nTokPos = rInStm.Tell();
    SvToken * pTok = rInStm.GetToken_Next();
    if( pTok->Is( pName ) )
    {
        BOOL bOk = FALSE;
        if( rInStm.Read( '=' ) )
        {
            pTok = rInStm.GetToken_Next();
            if( pTok->IsInteger() )
            {
                ULONG n = pTok->GetNumber();
                if ( n >= nMin && n <= nMax )
                {
                    *pValue = n;
                    bOK = TRUE;
                }
            }
        }

        if( bOk )
            return TRUE;
    }

    rInStm.Seek( nTokPos );
    return FALSE;
}

Now the method does much less than the old one, but everything needed at the
only place where it is used.

Code works in all tested modules. 

Please verify again.

yes, final version looks good---if n at calling site is changed from long to
ULONG and code is actually committed ;)

Of course. :-)
I will now push it to the outgoing repository.

.