Zachary Ware suggested I create an issue to discuss this:

I've started a worker using the worker name monson-win-arm32 and the password provided.

I think it is waiting for a change, there were no errors, but it didn't print anything.

Also, I don’t see anything in the list of builders that looks like it would be windows arm32, and it's not showing in the list of workers.

I'm looking at tools/buildbot/test.bat and it seems like it might be a good place to use SSH to run the test on arm32 device, but I'm not clear on where it is called from or what the best way to detect that project is being cross-compiled.  

Should I add an "-arm32" switch here?

New changeset 67ff6a103a1184b572750c30e231968c963e72f8 by Steve Dower (Paul Monson) in branch 'master':
bpo-36511: Windows ARM32 buildbot changes (GH-12917)
https://github.com/python/cpython/commit/67ff6a103a1184b572750c30e231968c963e72f8

Going to leave this in commit review for a bit, in case hitting merge suddenly flushes out some reviews :)

New changeset 4f820723c86c94f857d8d8de47a3c28f985946bd by Miss Islington (bot) (Paul Monson) in branch 'master':
bpo-36511: Windows arm32 buildbot changes (remove extra space) (GH-13351)
https://github.com/python/cpython/commit/4f820723c86c94f857d8d8de47a3c28f985946bd

Paul - looks like there's a timeout being hit due to lack of output. Any ideas?

https://buildbot.python.org/all/#/builders/203/builds/6

I've been investigating issues with test failures on my Windows buildbots seemingly not showing up in the master's web interface (but just showing warnings), and it appears likely due to this change.

For example, test_urllib (a test problem from issue 36948) was failing but only appeared as a warning (the first such case on my Win10 builder for example is https://buildbot.python.org/all/#/builders/3/builds/2661).

This change appears to force the exit code of test.bat to be successful (0) regardless of the results of the test run, thus obscuring any test failures from a parent process such as the buildbot.  As an aside, it also looks like any failures along the new arm32ssh branch would be obscured, assuming that the ssh operation would otherwise have propagated the remote exit code.

I think just leaving off the exit code (so just "exit /b") should propagate the rt.bat result properly.  Or, a goto to a label at the end of the file and letting it exit naturally would mimic the prior behavior.

I did a quick test and it looks like exit /b %ERRORLEVEL% will propagate the exit code.

Yeah, I think you're right.

It looks like without an explicit code, it won't propagate the result as the exit code of cmd itself for those cases where cmd does exit (which would include the buildbots).

New changeset 51394b8c3d42e6e6d368251ff6f0612495724fc0 by Steve Dower (Paul Monson) in branch 'master':
bpo-36511: Ensure error code propagates out of batch files (GH-13529)
https://github.com/python/cpython/commit/51394b8c3d42e6e6d368251ff6f0612495724fc0

New changeset f355069a3337711642c3403429afb9faef93f512 by Steve Dower (Paul Monson) in branch 'master':
bpo-36511: Add buildbot scripts and fix tests for Windows ARM32 buildbot (GH-13454)
https://github.com/python/cpython/commit/f355069a3337711642c3403429afb9faef93f512

New changeset a1952122a3a20272e4db9e3c6205009a91c73ccf by Steve Dower (Paul Monson) in branch 'master':
bpo-36511: Improve ARM32 buildbot scripts (GH-14251)
https://github.com/python/cpython/commit/a1952122a3a20272e4db9e3c6205009a91c73ccf

New changeset 43615485f1222cd9a94421a96cb7110bb52e1b8b by Steve Dower in branch '3.8':
bpo-36511: Improve ARM32 buildbot scripts (GH-14251)
https://github.com/python/cpython/commit/43615485f1222cd9a94421a96cb7110bb52e1b8b

New changeset f8dd77d36067fd7be614edde1e5e9e7467c450dc by Steve Dower (Paul Monson) in branch 'master':
bpo-36511: Fix -u parameters for ARM32 tests (GH-14280)
https://github.com/python/cpython/commit/f8dd77d36067fd7be614edde1e5e9e7467c450dc

New changeset f3e38ec7f014557296f6cc7b60a33d65faad1716 by Steve Dower in branch '3.8':
bpo-36511: Fix -u parameters for ARM32 tests (GH-14280)
https://github.com/python/cpython/commit/f3e38ec7f014557296f6cc7b60a33d65faad1716

New changeset ed70a344b5fbddea85726ebc1964ee0cfdef9c40 by Steve Dower (Paul Monson) in branch 'master':
bpo-36511: Fix failures in Windows ARM32 buildbot (GH-15181)
https://github.com/python/cpython/commit/ed70a344b5fbddea85726ebc1964ee0cfdef9c40

New changeset d85c5670ff1611202a25c9e0967148e72c114de9 by Victor Stinner in branch '3.8':
[3.8] bpo-37531: Fix regrtest timeout for subprocesses (GH-15072) (GH-15279)
https://github.com/python/cpython/commit/d85c5670ff1611202a25c9e0967148e72c114de9

New changeset ed00fee7d4c28489b5d6e4580b1e1fe0855687ca by Miss Islington (bot) in branch '3.8':
bpo-36511: Fix failures in Windows ARM32 buildbot (GH-15181)
https://github.com/python/cpython/commit/ed00fee7d4c28489b5d6e4580b1e1fe0855687ca

I manually restarted and merged the stalled 3.8 backport for 3.8.0b4.  Can this issue now be closed?

> Can this issue now be closed?

I spoke too soon. It appears there's another open backport PR-14244 for this issue that needs some conflict resolution to be merged.

New changeset 55d12ce8b8d397dd4e376bb0d1c27b3cced0ef50 by Zachary Ware (Paul Monson) in branch 'master':
bpo-36511: clean up python process before deploy on ARM Windows buildbots (GH-14431)
https://github.com/python/cpython/commit/55d12ce8b8d397dd4e376bb0d1c27b3cced0ef50

I rejected PR 14244 because there are now many conflicts: the commit f355069a3337711642c3403429afb9faef93f512 should be backported manually to Python 3.8.