Master failure

[mulkieran]

Related: #530.
Related: #776.
Related: #766.

[mulkieran]

Very wierd Jenkins failure: https://stratis-jenkins.rhev-ci-vms.eng.rdu2.redhat.com/job/stratisd/1375/.

Error where actual blockdev size is 1/10 the size of the recorded block dev size. I'm going to restart Jenkins and see what happens.
However, this is a definite place where error reporting could be better. If we got the blockdev size, we had a handle on the blockdev itself, for example, its devnode, and this could have been included.

[mulkieran]

Blocked by #1004.

[mulkieran]

No longer blocked.

[mulkieran]

It's been a month and no feedback, so I'm moving it back to in-progress and seeing what more I can do w/ it.

[mulkieran]

rebased.

[mulkieran]

I'm putting this into Stratis 1.1 preliminary review again as we may consider the problem of identifying the context of errors more urgent. We might want to consider a restricted goal, which does not attempt chaining, but does use failure crate to capture stack trace at point of error.

If we did that, we could perhaps start by sticking in failure errors only at leaf errors, i.e., locations where the error is picked up from an external crate.

[mulkieran]

There is another possibility worth exploring. Perhaps we can simply make our existing error types implement Fail and then we can get the backtrace sort of for free.

[mulkieran]

I'm taking this off in-progress, because it is now one of several alternatives that I'm exploring.

[mulkieran]

rebased. Keeping around for reference only.

[mulkieran]

Superceded by #1499. We can always take a look at this as a reference if need to, but might as well close at this time.