Fs cleanups

[joshmoore]

A few minor things I found while testing open tickets (7856, 11262):

• Remove unused mention of log4j's MDC
• On import failure, raise an exception rather than permit a NPE.
• Enable more use of the server-side memoizer cache.

Of these, the last is the only that one that really requires re-testing. While importing SDT files, each fileset log had at least one instance like this:

/ManagedRepository/root_0/2013-12/05$ grep "no directory given" *.log
07-38-49.111.log:2013-12-05 07:38:52,827 DEBUG [                   loci.formats.Memoizer] (3-thread-3) skipping memo: no directory given

meaning that multiple calls to SDTReader.setId were being made. For any file format which takes a substantial amount of time on the setId, this could be substantial large savings.

/cc @melissalinkert, @pwalczysko, @jburel

--no-rebase

[snoopycrimecop]

Conflicting PR.Removed from build OMERO-merge-develop#511. See the console output for more details.

[joshmoore]

Merging gh-1866 and rebasing.

[ximenesuk]

So the SDTs in test_images_good import okay as far as the logging goes.

However, all for images appear black and two have no channels at all (but they do have T and t dimensions). Is this expected or should I be seeing something?

I tested this including the final commit. Is there a way to test that commit specifically? Setting the wait very high or low?

[joshmoore]

@ximenesuk : for testing e83548e:

• set memoizer_wait to 0 in which case there should be no memoizer files created.
• set it to something very long, and then memoizer files should even be created for instant fakes.

[joshmoore]

As for the SDTs, they are fairly unrelated to this PR. Any import (with a long setId time) would suffice.

[joshmoore]

@ximenesuk didn't see any large improvements here for 10740. Instead, I'll reduce the number of transactions in gh-1885. Does anyone see any issues with this PR as it stands? @ximenesuk , @pwalczysko ?

[ximenesuk]

I don't see any issues.