Panic: TestProvisioningMachinesWithSpacesSuccess

[Curtis Hovey]

As seen in
    http://reports.vapour.ws/releases/issue/56969ab8749a564e0efe8420

TestProvisioningMachinesWithSpacesSuccess consistently panics in trusty ppc64el unit tests.
The host is using the latest gccgo-go 1.2.1-0ubuntu1.2

[Michael Hudson-Doyle (mwhudson)]

I can reproduce this on the porter box on the controller-rename branch, but not on master (which in some ways is a relief: there was a possibility that this was a regression caused by the new gccgo-4.9 that migrated into trusty-updates a day or two ago). Digging a little more now.

[Michael Hudson-Doyle (mwhudson)]

The immediate failure is trying to print a string which claims to be 8026363837385569385 bytes long, which is a pretty suspicious number. In fact:

>>> struct.pack('L', 8026363837385569385)
'ithub_co'

So basically something that should be a pointer to a go string is pointing to inside string data somewhere. This makes me think of https://bugs.launchpad.net/juju-core/+bug/1517611 especially because the tests pass on wily.

https://wiki.ubuntu.com/MichaelHudsonDoyle/Go15InTrusty pls

[Michael Hudson-Doyle (mwhudson)]

This (absurd) patch fixes this problem in my testing: http://paste.ubuntu.com/14491526/ leading me to believe it is the same bug as before.

[Curtis Hovey (sinzui)]

this issue is now visibly in controller-rename and 1.25. 1.25 has not been getting the same changes as controller-rename. The gccgo-go package was updated recently an could be the cause.

[Curtis Hovey (sinzui)]

2 days ago the trusty host had
   go version xgcc (Ubuntu 4.9.1-0ubuntu1) 4.9.1 linux/ppc64le
today it has
   go version xgcc (Ubuntu 4.9.3-0ubuntu4) 4.9.3 linux/ppc64le