I received from feedback from upstream on this.
On Tue, Feb 13, 2018 at 12:39:17AM -0500, Roberto C. Sánchez wrote:
> - Is it correct to be concerned about the ABI in this case? (I know
> that changing type/order of parameters and struct members would break
> ABI compatibility, but I am not as certain about return type).
> - Assuming that something needs to be done to preserve ABI compatiblity,
> is Olly's recommended approach appropriate/viable? (It seems so to me)
> - Assuming the approach is viable, is my implementation correct?
Bob Friesenhahn pointed out that changing return type from void to int
does not change the ABI. Additionally, the prototype for the function
in question is not visible to consumers of the graphicsmagick library.
Based on the feedback, I have dropped the third patch and am proceeding
with an update that includes only the two backported patches from
Roberto C. Sánchez