Re: Tiny patch for dib.c

[Mike McCormack]

[e-mail] wrote:
>  if (colors > 256)
> - {
> -
>    ERR("called with >256 colors!\n");
> -
>    return NULL;
> - }
> -
>    WARN("called with >256 (%d) colors!\n", colors);

We appear to only get down this codepath if bmBitsPixel <= 8, so there 
should be no more than 2^8 colours.

Maybe you can investigate how we can get a DIB with bmBitsPixel <= 8 and 
(biClrUsed > 256 or bcBitCount>8) ?

Mike

Tiny patch for dib.c

On Thu, May 04, 2006 at 12:16:10PM +0900, Mike McCormack wrote:
> Maybe you can investigate how we can get a DIB with bmBitsPixel <= 8
> and (biClrUsed > 256 or bcBitCount>8) ?

I'm not an expert on this, but I supposed the application called it with
an off-by-one value of 257?  Do you think the problem could be with
Wine?

Leslie

Tiny patch for dib.c

[Dmitry Timoshkov]

[e-mail] wrote:
> On Thu, May 04, 2006 at 12:16:10PM +0900, Mike McCormack wrote:
> > Maybe you can investigate how we can get a DIB with bmBitsPixel <= 8
> > and (biClrUsed > 256 or bcBitCount>8) ?
>
> I'm not an expert on this, but I supposed the application called it with
> an off-by-one value of 257?  Do you think the problem could be with
> Wine?

Most likely the 'color' should be silently truncated to 256, we had same
kind of the problem with X11DRV_[Set|Get]DIBColorTable.

Tiny patch for dib.c

[Francois Gouget]

On Thu, 4 May 2006, [e-mail] wrote:
> On Thu, May 04, 2006 at 12:16:10PM +0900, Mike McCormack wrote:
> > Maybe you can investigate how we can get a DIB with bmBitsPixel <= 8
> > and (biClrUsed > 256 or bcBitCount>8) ?
>
> I'm not an expert on this, but I supposed the application called it with
> an off-by-one value of 257?  Do you think the problem could be with
> Wine?

I think Mike's idea is to check whether it could be a Wine bug that 
causes the application to pass us an invalid value (e.g. some API 
returning 256 instead of 255 or some such). This happens sometimes and 
then we would want to fix the original bug.

However, as Dmitry said, this might also be a simple application bug. If 
you can add a conformance test to check how Windows behaves in this 
situation it could also help clarify things (apologies if your patch 
already contained such a test).