Investigate thread safety

[tbirdso]

From #44 (comment):

When I run tests in parallel with ctest -j8, I get:

  9 - IOOMEZarrNGFF_readSubregion (Subprocess aborted)

Serially, they pass.

I have been able to reproduce these errors with the -j<n> flag with different tests failing. In particular I've noticed errors where Tensorstore appears to fail to read different image dimensions. There may be an issue with store contexts not being unique among threads or processes.

Tasks

• Add a test to verify that using two OMEZarrNGFFImageIO instances with two ImageReaders of different dimensions in sequence does not result in an error + metadata is read correctly
• Investigate failures in running tests in parallel
• Add a test to verify that running multiple threads with one OMEZarrNGFFImageIO instance and ImageReader per thread with different data sources / regions does not result in failures
• If thread/process/context safety issues are identified, start by documenting the issue ("multiple instances are not allowed") or seeing if we can enforce one context at a time.
• If safety issues are identified, revisit and determine how/whether we can refactor for thread/process/context safety.

[tbirdso]

fyi @thewtex @dzenanz

[dzenanz]

This is a good identification of what to look at next.

[dzenanz]

After all the contents are added to in-memory zip file, the buffer size is 47447. After closeZip() is called, the buffer size increases to the correct value of 47685. This leads me to suspect that destructor is not invoked sometimes.

I will now try to fix this.

[tbirdso]

Thanks @dzenanz , that is in reference to the in-memory zip issue correct? #36

[dzenanz]

Yes.

[dzenanz]

I now realize that this discussion should be moved to that issue.