u1sdtool is hanging in current precise

u1sdtool doesn't appear to continue running, but doesn't return the shell prompt at all when done in this way now.

^C won't even exit whatever is going on here. no process named u1sdtool remains running either.

python is still running when the process is stopped via ^Z. investigating.

hung process doesn't respond to SIGINT from kill either and must be sent SIGKILL

quick poke with pdb doesn't turn up obvious cause of the hang, but the hang can be duplicated within pdb.

This is an issue in the twisted package. Another small change that happened prior to the gireactor/gtk3reactor landing, is needed for it to work properly, and I wasn't aware of it in the backport.

I would just like to note that this is not the first time that an untested backport of a Twisted patch was deployed on Ubuntu or Debian and caused problems.

On the Twisted project, we have an elaborate continuous integration system (see <http://buildbot.twistedmatrix.com/>) that verifies that our software is correct, as well as pre-release testing with our user community, many of whom have similar testing rigs for integrating Twisted with their application. We really need that infrastructure because Twisted's whole job is managing the numerous subtle interactions with underlying operating system interfaces, including signals and threads, pipes, sockets, etc etc.

It seems like when a distribution patches something, the patches are widely distributed without any similar testing process, which means that we get hard-to-debug issues like this one, that are the result of dependent changes not being present. The fact that this one got caught before release is great, but others have slipped through into releases (if I recall correctly, even LTS releases) in the past.

In the future, it would be great if Ubuntu could work with us to do point-releases that get vetted the same way, rather than shipping untested patches.

Thanks, Glyph, for the feedback. I will raise this with some of the engineering managers at Canonical and see if there is a way we can help to improve this.

> I would just like to note that this is not the first time that an untested
> backport of a Twisted patch was deployed on Ubuntu or Debian and
> caused problems.

note that this patch was deployed to the development version of Ubuntu only.

> note that this patch was deployed to the development version of Ubuntu only.

I did note it - that's what I meant by "this one got caught before release". Maybe that means things are already getting better! :-).

However, it did still get caught by testing of u1sdtool, not by testing Twisted, so I think the current process is still risky for features not used by u1.

This bug was fixed in the package twisted - 11.1.0-1ubuntu2

---------------
twisted (11.1.0-1ubuntu2) precise; urgency=low

  * debian/patches/01_posix_wakeups.patch:
    - Backport wakeup change from trunk gtk2refactoring (LP: #935756)
 -- Rodney Dawes <email address hidden> Thu, 08 Mar 2012 11:25:48 -0500