[Swan-dev] test system run: corrupt charon log files

[Andrew Cagney]

On 20 July 2016 at 18:21, D. Hugh Redelmeier <hugh at mimosa.com> wrote:
> Does anyone know what's going on here?

One cause of crud in a file is it being left open/write during a shutdown.

Was strongswan cleanly shutdown by final or left running?

> I happened to look at
>   interop-ikev1-strongswan-02-psk-responder/OUTPUT/east.charon.log
> in my most recent run of the tests.
>
> The file seeems to be corrupt.  It does not start at the beginning of
> the log (judging from this log in a previous run).  It ends with a
> large sequence of NUL bytes and then a few lines of text.  The NULs
> start about 51k into a 464K file.
>
> The previous run (with a different version of strongswan) produced a
> log of similar size but without a pile of NULs.
>
> interop-ikev1-strongswan-03-psk-initiator/OUTPUT/west.charon.log seems
> to be similary corrupt.  Again, the previous run's west.charon.log
> doesn't seem to have this problem.
>
> This run was testing HEAD as of July 15, with one change:
> In testing/libvirt/fedora22base.ks lines from "# Extra yum stuff" on were
> deleted (except for the "%end" at the end).
>
> This test run used strongswan-5.4.0-2.fc22.x86_64.
> The previous test run used strongSwan 5.3.2

I don't think it is strongswan - I also see lots of nulls and don't
have your change.
The hosts are Fedora23.

For instance: http://testing.libreswan.org/results/testing/2016-07-20-0106-3.18dr3-171-g67ab67f-master/interop-ikev2-strongswan-24-aes_ctr/OUTPUT/east.charon.log
looking at final.sh, it doesn't appear to shutdown strongswan.

> The test system considered this test to have passed, with whitespace
> differences.  But then I don't think that the charon log is considered.
>
> Is this just expected flakiness of the test system or is it something
> else?