MinGW trunk build available on Appveyor
As many of you know, I have been building MinGW for quite a while, and posting results at Ruby MinGW Test Results.
I recently added a 64-bit build to BinTray, along the three custom MSYS2/MinGW packages I use. Log files for every build are also posted, and all archives containing code are signed. I did so so I or anyone else can use the build for testing on Appveyor.
Since the packages may change (OpenSSL 1.1.0f will change to 1.1.0g sometime), the cmd script for loading the packages is contained in the ruby_trunk.7z file. An example of its use is at appveyor MSP-Greg/appveyor-ruby with the script at MSP-Greg/appveyor_ruby.
Some people are against using a build that is not cloud based. For the time being, I have chosen to do so for the following reasons:
Appveyor issues - see build 1.0.4718 for an example.
Parallel test issues - I still have problems with issues running parallel.
Intermittent Failures - see build 1.0.4678. Once in a while, I see this failure. Always, run the test (parallel) again, and it passes.
Environment issues - there is one test I can run successfully in a straight windows env, but running in a hybrid env (with some *nix commands available), it fails.
Two other Windows/MinGW fails, both of which I opened issues on. See above mentioned results page for info.
Skips aren't fails. I would prefer to see if skips increase dramatically between builds.
If there are new failures/errors, I can easily isolate the them, run non-parallel, run in a make env, run in a 'ruby env', etc.
Lastly, two items:
At present, RubyInstaller2 builds (2.4.1 and snapshot) do not run
test-spec. Also, MSYS2/MinGW updated to gcc 7.1.0 recently (appveyor still has 6.3.0), and I have test issues when using 7.1.0. For the time being, I'm using 6.3.0.
I expect to be doing builds at least five times a week, maybe more.