*** Bug 62343 has been marked as a duplicate of this bug. ***

Here is a simple fix:

Created attachment 5824 [details] [review]
the patch

This patch seems to be already applied.

Also, since it's patching the BMP code, and the problem is with
the XBM handler, I don't see how it will help :-/

Here is the problem code (context is allowed to be NULL on entry):

static GdkPixbuf *
gdk_pixbuf__xbm_image_load_real (FILE *f, XBMData *context, GError
**error)
{
	guint w, h;
	int x_hot, y_hot;
	guchar *data, *ptr;
	guchar *pixels;
	guint row_stride;
	int x, y;
	int reg = 0; /* Quiet compiler */
	int bits;

	GdkPixbuf *pixbuf;

	if (!read_bitmap_file_data (f, &w, &h, &data, &x_hot, &y_hot)) {
                g_set_error (error,
                             GDK_PIXBUF_ERROR,
                             GDK_PIXBUF_ERROR_CORRUPT_IMAGE,
                             _("Invalid XBM file: %s"),
                             context->tempname);
		return NULL;
	}

'context' is used in g_set_error, even though it may be NULL.

I attached the wrong patch. How could this happen ?!

Created attachment 5830 [details] [review]
the right patch

Somehow xbm really escapes me. I guess rm -f *.diff is in order.

Somehow xbm really escapes me. I guess rm -f *.diff is in order.

Created attachment 5831 [details] [review]
this patch is even more on topic

Looks fine to me. (There is an outstanding win32 patch
that changes the handling of context->tempnam a little
bit, but I don't think it interacts.)