Small additions to cheritest

[nwf]

And some tweaks.

Extracted from the long-lived caprevoke branch. No reason, I think, that these can't land in the tree independent of that work, tho'.

[nwf]

Unclear why CI marked this as failed; Jenkins itself claims "passed but marked unstable".

[jrtc27]

For example:

Timeout running cheritest: timeout running /bin/cheriabitest -a -x > /tmp/cheriabitest.xml

which doesn't show up in the Tests tab since that results in no JUnit file at all.

[jrtc27]

Probably raced and sleeping in sigsuspend, never waking up again?

[jrtc27]

How are you ever expecting to wake up, anyway? What's delivering that signal?

[jrtc27]

Hm, I expect the idea is for the close(pfd[1]); to cause an EOF condition on the pending AIO read? Strange that it consistently works for cheritest but not cheriabitest. I wonder if one of the sig* syscalls did the wrong thing (and perhaps even reported an error that isn't checked).

[nwf]

I just added CHERITEST_VERIFY around all the sig* calls and, sadly, they all appear to be claiming that the right thing happened in both cheritest and cheriabitest. Other than that failure, I believe this is good to go?

[jrtc27]

We need to figure out what's going on with AIO...

[jrtc27]

I believe we added tests to cheribsdtest along these lines a while ago in dev