On Jun 19, 2008, at 14:13, Brian Aker wrote:
On Jun 19, 2008, at 12:58 PM, Brad Fitzpatrick wrote:
But if we want to include libmemcache with memcached (at least for
testing), we could just as well write the tests in C too. *shrug*
I would be fine with that.
Would this require (or allow for) any unification of coding standards?
I've got dozens of tests and once the binary bits are complete we
can flip/flop the protocols and see that they are returning the same
results.
I run a fairly comprehensive test four times during my client build.
Once for each combination of ipv4/ipv6 and text/binary.
I wouldn't recommend bundling my client with memcached, but the test
strategy is pretty good.
We could do something with the buildbots to pull the tests each new
build of memcached.
I *do* think it would be good to have a builder running that would
play other people's tests on each build (sandboxed or something).
I've got tests that range from a text file you pipe through netcat and
see if the server crashes to all the junk my java client does as it
tries to test itself.
--
Dustin Sallings
