From: merch-redmine@... Date: 2020-08-31T20:22:29+00:00 Subject: [ruby-core:99806] [Ruby master Bug#15209] r64925 order-only-prerequisites - make targets - breaking change Issue #15209 has been updated by jeremyevans0 (Jeremy Evans). Status changed from Open to Closed r64925 was reverted shortly after this in commit:3757515501347685bfbd9ef11bc03e93c66d1dc9, so this can be closed. ---------------------------------------- Bug #15209: r64925 order-only-prerequisites - make targets - breaking change https://bugs.ruby-lang.org/issues/15209#change-87325 * Author: MSP-Greg (Greg L) * Status: Closed * Priority: Normal * Backport: 2.3: UNKNOWN, 2.4: UNKNOWN, 2.5: UNKNOWN ---------------------------------------- This commit has changed behavior of make targets such that make install or make install-nodoc is now running tests. Viewing Travis history shows that before the commit trunk builds took approx 17 - 23 minutes, they are now taking up to 49 minutes. If you view the log for Travis Job #27742.2: https://api.travis-ci.org/v3/job/437903711/log.txt First off, it's 222k lines. test-all is being done three times, test-spec is being done twice. I'm not particular about target naming, but I think targets simiar in functionality to what was available before r64925 should continue to be available. Since MinGW has been added to CI, ruby-loco can be focused on whether a correct install is being built/packaged. All its tests are run without make on install. All tests are going to separate log files (one each for STDOUT & STDERR), and the tests are also wrapped in a timer such that, regardless of whether one suite freezes/locks, all tests are run. Because of this, I need separate targets for each. JFYI, most of the changes needed to run on install were in RubyGems & mkmf. Of course, I'm still interested in consistent, reliable, and repeatable testing. Random is also a goal... Thanks, Greg -- https://bugs.ruby-lang.org/ Unsubscribe: