Added bounds test to prevent crash on badly formed E00 ARC lines

[gaige]

There are conditions with some files that are badly formatted which will cause some of these numbers to not be positive. This is at least a rudimentary test for failure. Of course, out of bounds in the other direction isn't tested, but it was unclear where and how to do that.

Unfortunately, I cannot locate the file that caused this problem in the wild. However, these conditions can occur and are fatal.

ogr/ogrsf_frmts/avc/avc_e00parse.c

720a721,727
>             // we have to have at least 2 vertices for an "arc", unless converage ID is <1
>             if ((psArc->numVertices<0)||(psArc->nRPoly<0)|| (psArc->nLPoly<0)||(psArc->nTNode<0)||(psArc->nFNode<0)) {
>                 CPLError(CE_Failure, CPLE_AppDefined, 
>                          "Error parsing E00 ARC line: \"%s\"", pszLine);
>                 return NULL;
>             }
>             

[Even Rouault]

Gaige,

the normal process for changing files belonging to AVCBIN/AVCE00 is to submit them to upstream AVCE00. CC'ing Daniel Morissette, the AVCE00 maintener.

[Jukka Rahkonen]

I think that this patch has not been applied. Dmorissette?

[Even Rouault]

Removing obsolete milestone