[issue32823] Regression in test -j behavior and time in 3.7.0b1

Terry J. Reedy added the comment: Thanks for checking on this. 3.7.2 and 3.8.0a2 64 bit installed, as well as respository debug 32 bit 3.8.0a2+, are back to normal, with the quickest tests reporting within a few seconds. I will assume fixed for Zach also, hence closing. --

[issue32823] Regression in test -j behavior and time in 3.7.0b1

Cheryl Sabella added the comment: Terry, Do you still see this happening? When I run the tests on Windows 10 with 12 CPUs (using -j0), the tests run quickly with just the last one taking more time. 0:01:42 [418/420] test_socket passed (40 sec 239 ms) -- running: test_multiprocessing_spawn

[issue32823] Regression in test -j behavior and time in 3.7.0b1

Zachary Ware added the comment: Ok, interesting. The above results are from running on my Windows 8.1 buildbot, running on Azure. Running on a local fully-updated (I think) Windows 10 VM, I get results similar to what Terry reports: results come in batches of X for

[issue32823] Regression in test -j behavior and time in 3.7.0b1

Zachary Ware added the comment: I can't reproduce this. Even running `python -m test -j14` on a machine with 2 vCPUs, I get my first results at 2s, and even test___all__ passes in only 6 (though on my first run, with -j0 (== -j4), test___all__ took 41s but other

[issue32823] Regression in test -j behavior and time in 3.7.0b1

Terry J. Reedy added the comment: python -m test -j0 -- ___ Python tracker ___ ___

[issue32823] Regression in test -j behavior and time in 3.7.0b1

Zachary Ware added the comment: Terry, could you give me the full command you use to get that result? -- ___ Python tracker ___

[issue32823] Regression in test -j behavior and time in 3.7.0b1

Terry J. Reedy added the comment: Yes. For 3.7 and 3.8, nothing appears until 41 seconds because test___all__ takes 40 seconds. For 3.6, the batch result appears at 37 seconds because test___all__ runs in 36 seconds. -- ___

[issue32823] Regression in test -j behavior and time in 3.7.0b1

Ned Deily added the comment: Terry, is this still an issue? Are other Windows developers seeing this? -- nosy: +ned.deily, paul.moore, steve.dower, tim.golden, zach.ware ___ Python tracker

[issue32823] Regression in test -j behavior and time in 3.7.0b1

Terry J. Reedy added the comment: For me, on Windows, the tests for time, datetime, and _strptime have failures on the master branch, so I don't want to submit a PR just now. -- ___ Python tracker

[issue32823] Regression in test -j behavior and time in 3.7.0b1

Change by Terry J. Reedy : -- Removed message: https://bugs.python.org/msg312021 ___ Python tracker ___

[issue32823] Regression in test -j behavior and time in 3.7.0b1

New submission from Terry J. Reedy : For me, on Windows, recent changes, since 3.7.0a4, have greatly reduced the benefit of j0 (= j14 on my system). For 3.6.4 (installed), the times are around 13:30 and 2:20. The times for 3.7.0a4 (installed) are about the same parallel and