Hi,
On Fri, May 15, 2015 at 11:06:18PM -0700, Afif Elghraoui wrote:
OK, feel free to discuss the diff here in case you might be in doubt.
I'm not sure how useful this patch of kazlib (attached) is for
forwarding. It seems to me too much like a workaround. According to
the comments and what I see in the diff, all it does is make the
data structure work with a pointer rather than the object itself
(also represented by a pointer). The stated objective of this is to
maintain compatibility with qsort, but I don't know why the result
could not have been taken as is and passed as an address to qsort.
This also seems like something that would break the code if it is
not included, so I'm not sure how tests would have passed using the
Debian packaged version of kazlib. I should probably check the build
logs for anything suspicious that may have been ignored.
I think the next step would be to bring this up with kmer upstream
and consider patching the kmer source to work with kazlib the way it
is. What do you think? Offhand I don't think patching kmer to
achieve the same ends would be very difficult.
I personally would try to convince kmer upstream to adapt to official
kazlib.