On Tue, Nov 11, 2014 at 10:11 AM, Marko Tiikkaja <marko@joh.to> wrote:
On 2014-11-11 18:52, Jeff Janes wrote:
The problem is that mdc_read is getting invoked with an len argument of 0. When it successfully reads 0 bytes, it interprets that as an error. I can't figure out why it is getting invoked with a length of 0, too many pointer indirections for me to follow at this time.
This sounds a lot like another bug I've seen. Can you try this patch?
That fixes it. I've tested all lengths from 1 to 133,000
I've added a regression test to it, in the attached.
How do we go about incorporating it, as far as bumping the extension version-number and such? Or does it not get bumped at all because there is no change to the interface?