update on (delayed) 1.10.1 release

[Brian Warner]

Hey folks..

Just wanted to give y'all an update on the 1.10.1 release. We haven't
shipped a beta yet. The scope of the work is still the same (we haven't
added anything significant to the milestone), but everybody involved has
been swamped with other responsibilities, so the work has been going
slowly.

It *has* been moving forward, however, and we've managed to finish the
majority of the code for the remaining tickets. Here's the list of
what's left:

 https://tahoe-lafs.org/trac/tahoe-lafs/query?status=assigned&status=new&status=reopened&group=status&milestone=1.10.1

* #2329 "cp -r stops with an exception":

We've performed a massive analysis of all the different ways you might
run "tahoe cp" (from=local, to=tahoe), how they behave under 1.10.0,
current trunk, and regular unix /bin/cp . We decided on a table of ideal
behaviors, built an extensive unit test to check them, and rewrote the
code (in tahoe_cp.py) to implement the table. All that's left is to
update the docs to explain the new behaviors.

* #2388 "AssertionError: paths must be Unicode strings, when starting a
node"

This turned out to be a serious regression that basically broke the SFTP
server entirely. The fix is pretty easy, though, and we've got a patch
awaiting review to fix it.

* #2312 "DeprecationWarning about passing non-bytes header values"

We fixed the ones that we could, and have identified the ones that we
can't (because they live in non-Tahoe code). We may bump this to 1.11,
as it only causes some noisy tests, and only under python2.6.

* #2034 "Update for mutable files is sensitive to the number of servers"

We landed a patch, and it probably fixes the problem, but now we'd like
to add a test too.

* #2375 "make tests work under a non-ASCII path"

There's an (incomplete?) patch in development for this one.

* #2215 "mitigate heartbleed vulnerability"

We've discussed it a lot, and it'd be nice if Tahoe could refuse to run
if it detects that it's using a vulnerable version of OpenSSL, but it's
looking difficult, and we may kick it down the road into 1.11

* #195, #1980, #1981, #2363 (windows packaging)

We've gotten a new windows buildslave up and running (although not yet
passing tests), which will make these easier. I've also ressurected the
flappserver-based tarball upload tool, so the generated installers/eggs
can be put somewhere useful. Apart from these, I think there's still a
lot of work to be done.


That's everything on the plate for 1.10.1 . There are some
infrastructure things that have been done too (buildbot cleanup, OS-X
packaging). But things are happening.

I don't have a strong timeline for when we'll cut the beta, but I'm
hoping to do it before PyCon (mid-April).

cheers,
 -Brian