In an ideal world, we'd have test cases for each bug ever found. In an ideal world, we'd only call something a regression when we have a test case for it. In an ideal world, we'd have tireless testers with infinite time to get things right. In an ideal world, we do not have to compromise.
In the real world, we will always have bugs in every release. In the real world, we should never release something which shows bugs in our test suite. In the real world, if we find bugs without test coverage, we have to weigh things. In the real world, manual tests do not improve on the n-th iteration, as people get bored/tired/don't have the time. In this case, we know of a bug, but have no test case for it. In this case, it used to work before, but no longer. In this case, it is not affecting users of the built server. In this case, there are workarounds for people building httpd (which is a much smaller number than users). Therefore, I vote +1 on releasing 2.4.22. PS. I would appreciate if we could pour some of the energy visible here into having a build server that just runs some build configurations nightly/weekly. 2.4.19 failed because of Windows issues, 2.4.22 has hick ups as of build configuration issues. Only automation will improve things for the releases to come.
