Re: Escher images

[Jody Goldberg <jody gnome org>]

On Fri, Jan 16, 2004 at 02:55:54PM +0000, Nick Lamb wrote:
> It looks to me like the check on opcode for BIFF records in an Escher
> container is overly strict. I have a file here which contains a chart
> with a massive JPEG background. Current 1.2 CVS crashes while trying to
> handle that part of the document, but by adding BIFF_CHART_gelframe to the
> list the JPEG is merely ignored (presumably not yet supported). Does
> that make sense? Obviously the crash when an opcode doesn't match should
> be fixed too, but I had no time to fully diagnose that.
>
> My test document is XOR encrypted, and therefore can't be opened without
> the XOR crypto patch (or its predecessor). The change needed to open this
> without crashing is inline at the bottom of this mail, it's a one liner.
> Unfortunately the document is not only encrypted, it's also confidential,
> so if someone needs to take a look (Jody?) they'll have to email me off
> list and agree to use it for research only and then destroy it.
>
>
> diff -u -r1.100 ms-escher.c
> --- ms-escher.c 14 Dec 2003 04:58:24 -0000      1.100
> +++ ms-escher.c 16 Jan 2004 14:46:07 -0000
> @@ -210,6 +210,7 @@
>                         if (q->opcode != BIFF_MS_O_DRAWING &&
>                             q->opcode != BIFF_MS_O_DRAWING_GROUP &&
>                             q->opcode != BIFF_MS_O_DRAWING_SELECTION &&
> +                           q->opcode != BIFF_CHART_gelframe &&
>                             q->opcode != BIFF_CONTINUE) {
>                                 g_warning ("Unexpected record type 0x%x @ 0x%x;", q->opcode, q->streamPos);
>                                 return NULL;

Good catch.  That area is fairly murky.  I've made a similar change
in the initial check too.   We probably don't need that.  I suspect
that it would be more correct for us to expect just one record type
that would be defined by the initial record.

eg all draws or all gelframes

I wonder if we even need CONTINUE ...

Applied to HEAD and 1.2.  It will be in 1.2.5 this weekend.