Gh-2711: conditionally run tests on windows

[lb324567]

Related Issue

• Resolve Add capability to run tests on Windows #2711

[lb324567]

Commit f96aa22 should pass (the tests anyway), as its only running the tests on Ubuntu (because of the keywords it's looking for).
Commit dd6a883 will probably fail because of issues with the tests running on windows, but does show that if the keywords are in the branch name the tests will run on windows.

[codecov-commenter]

Codecov Report

Merging #2716 (7398b99) into v2-alpha (8a43ebd) will decrease coverage by 0.20%.
The diff coverage is n/a.

@@              Coverage Diff               @@
##             v2-alpha    #2716      +/-   ##
==============================================
- Coverage       65.73%   65.53%   -0.21%     
+ Complexity       2435     2268     -167     
==============================================
  Files             862      611     -251     
  Lines           27251    19188    -8063     
  Branches         3126     2259     -867     
==============================================
- Hits            17914    12575    -5339     
+ Misses           7980     5564    -2416     
+ Partials         1357     1049     -308     

Impacted Files	Coverage Δ
...peration/mapper/generator/TextMapperGenerator.java
...chq/gaffer/flink/operation/handler/GafferSink.java
.../gchq/gaffer/mapstore/impl/GetElementsHandler.java
...ketches/quantiles/StringsSketchKryoSerializer.java
...edicate/user/FederatedGraphWriteUserPredicate.java
...byteEntity/ByteEntityAccumuloElementConverter.java
...java/uk/gov/gchq/gaffer/mapstore/impl/MapImpl.java
...rame/converter/property/impl/FreqMapConverter.java
...hq/gaffer/mapstore/impl/GetAdjacentIdsHandler.java
...ov/gchq/gaffer/accumulostore/utils/TableUtils.java
... and 242 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update a820490...7398b99. Read the comment docs.

[GCHQDeveloper314]

I've just raised #2717 to cover the remaining Windows test failures. However, because it is not a priority issue, I don't think we should run this new Windows variant of CI for releases yet.
The next release is not planned to include the Windows fixes. If this PR were merged ahead of that release, the CI would fail when run on the release. We don't really want the CI to routinely fail on releases as this is misleading.

Also, under #2717 there will be extra changes required to the CI for Windows only (to install appropriate Hadoop libraries). If the release trigger were to remain disabled for now, it could then be enabled when #2717 is fixed (as there would be no more failures after this point).

[t92549]

I've just raised #2717 to cover the remaining Windows test failures. However, because it is not a priority issue, I don't think we should run this new Windows variant of CI for releases yet. The next release is not planned to include the Windows fixes. If this PR were merged ahead of that release, the CI would fail when run on the release. We don't really want the CI to routinely fail on releases as this is misleading.

Also, under #2717 there will be extra changes required to the CI for Windows only (to install appropriate Hadoop libraries). If the release trigger were to remain disabled for now, it could then be enabled when #2717 is fixed (as there would be no more failures after this point).

So for now, this should only run when "windows" is in the branch name?

[GCHQDeveloper314]

So for now, this should only run when "windows" is in the branch name?

Yes that's the idea, this way we can create branches and PRs to work on the Windows problems, but the releases will continue to build as usual and will pass. Then once the existing Windows issues are fixed, we can enable it for releases as well.

From a practical perspective, we could have release-disabled-for-now instead of release so that there's only a minor edit required to enable, that way this PR could be merged pretty much as it is, instead of taking the release bit out.

[lb324567]

So for now, this should only run when "windows" is in the branch name?

Yes that's the idea, this way we can create branches and PRs to work on the Windows problems, but the releases will continue to build as usual and will pass. Then once the existing Windows issues are fixed, we can enable it for releases as well.

From a practical perspective, we could have release-disabled-for-now instead of release so that there's only a minor edit required to enable, that way this PR could be merged pretty much as it is, instead of taking the release bit out.

I've made the change to use release-disabled-for-now as you suggested :)

[GCHQDeveloper314]

Because the branch for this PR contains the word 'windows', the tests are run on Windows and they fail because Windows isn't fully supported yet. This PR is ok to be merged despite the CI showing a failure.