Release proposal: v1.4.2

[rvagg]

This is a quick one, will give it 24 hours or less. Mainly to fix #988, proposed change being in #994.

• [25da0742ee] - build: improve vcbuild.bat (Bert Belder) #998
• [b8310cbd3e] - build: reduce tarball size by 8-10% (Johan Bergström) #961
• [58a612ea9d] - deps: make node-gyp work with io.js (cjihrig) #990
• [2a2fe5c4f2] - deps: upgrade npm to 2.6.1 (Forrest L Norvell) #990
• [84ee2722a3] - doc: minor formatting fixes. (Tim Oxley) #996
• [cf0306cd71] - doc: update stability index (Chris Dickinson) #943
• [fb2439a699] - doc: add robertkowalski as collaborator (Robert Kowalski) #977
• [f83d380647] - doc: update os.markdown (Benjamin Gruenbaum) #976
• [ae7a23351f] - doc: add roadmap, i18n, tracing, evangelism WGs (Mikeal Rogers) #911
• [14174a95a5] - doc: document roadmap, workgroups (Mikeal Rogers)
• [865ee313cf] - doc: add tellnes as collaborator (Christian Tellnes) #973
• [01296923db] - doc: add mscdex as collaborator (Brian White) #972
• [675cffb33e] - http: don't confuse automatic headers for others (Christian Tellnes) #828
• [7887e119ed] - install: new performance counters provider guid (Russell Dempsey)
• [4d1fa2ca97] - src: add check for already defined macro NOMINMAX (Pavel Medvedev) #986
• [1ab7e80838] - tls: proxy handle.reading back to parent handle (Fedor Indutny) #1004
• [755461219d] - tls: fix typo handle._reading => handle.reading (Fedor Indutny) #994

[rvagg]

FYI: Jenkins is being a for us and getting different build machines to build on different commits, hiding test failures that crept in before 1.4.1. I also had lots of trouble even getting 1.4.1 out because of this but I think I may have sorted it out (never say never with Jenkins).

CI on v1.x as it is now: https://jenkins-iojs.nodesource.com/job/iojs+any-pr+multi/207/ - expecting to see Windows failures this time. Will try #994 next.

[rvagg]

And there it is! Human error after all, the test-ci and introduction of TAP changes in CI were causing havoc so I'd reverted the test runs to be just test-simple and test-message but only test-message is running on Windows. I'm not actually sure why since it's doing vcbuild.bat release nosign x64 test-simple test-message and that's supposed to work. Trying just vcbuild.bat release nosign x64 test now.

https://jenkins-iojs.nodesource.com/job/iojs+any-pr+multi/208/

[indutny]

Yay! :)

[rvagg]

2015-02-28, Version 1.4.2, @rvagg

Notable changes

• tls: A typo introduced in the TLSWrap changes in #840 only encountered as a bug on Windows was not caught by the io.js CI system due to problems with the Windows build script and the Windows CI slave configuration. Fixed in #994 & #1004. (Fedor Indutny)
• npm: Upgrade npm to 2.6.1. See npm CHANGELOG.md for details. Summary:
  • 8b98f0e
    #4471 npm outdated (and only npm outdated) now defaults to --depth=0. This also has the excellent but unexpected effect of making npm update -g work the way almost everyone wants it to. See the docs for
    --depth
    for the mildly confusing details. (@smikes)
  • aa79194
    #6565 Tweak peerDependency
    deprecation warning to include which peer dependency on which package is
    going to need to change. (@othiym23)
  • 5fa067f
    #7171 Tweak engineStrict
    deprecation warning to include which package.json is using it.
    (@othiym23)
• Add new collaborators:
  • Robert Kowalski (@robertkowalski)
  • Christian Vaagland Tellnes (@tellnes)
  • Brian White (@mscdex)

Known issues

• Windows support has some outstanding failures that have not been properly picked up by the io.js CI system due to a combination of factors including human, program and Jenkins errors. See #1005 for details & discussion. Expect these problems to be addressed ASAP.
• Surrogate pair in REPL can freeze terminal #690
• Not possible to build io.js as a static library #686
• process.send() is not synchronous as the docs suggest, a regression introduced in 1.0.2, see #760 and fix in #774
• Calling dns.setServers() while a DNS query is in progress can cause the process to crash on a failed assertion #894

Commits

[rvagg]

I might just do this today, give me a few hours to work on Jenkins to make sure it's working properly and I'll go ahead and do a release.

PLEASE DON'T INTRODUCE SEMVER-MINOR COMMITS UNTIL THIS IS RELEASED

[rvagg]

@indutny @piscisaureus the real problem atm is that Jenkins isn't getting a !0 exit code from vcbuild. Check this out: https://jenkins-iojs.nodesource.com/job/iojs+any-pr+multi/209/nodes=iojs-win2012r2/console - note the SUCCESS at the bottom but 79 errors. So it's always going to be blue. Any clues what that might be about? Full command on this is this atm: vcbuild.bat release nosign x64 test.

[piscisaureus]

@rvagg

Check this out: https://jenkins-iojs.nodesource.com/job/iojs+any-pr+multi/209/nodes=iojs-win2012r2/console -

But these look like genuine failures. Or am I missing something?

[rvagg]

@piscisaureus genuine failures but the build is blue (not red), look at the last line:

Finished: SUCCESS

It's not reporting back. I'm about to hop on one of these machines and investigate whether it's the use of -J for test.py that's doing this (current theory), if you have any better ideas then let me know. Basically with the current test setup we can have failures on Windows that you won't see unless you're looking at console output.

[piscisaureus]

@rvagg I'm guessing that the exit code is zero because after the tests, jslint runs and succeds.

Batch processing isn't aborted when a command returns a nonzero exit code, So even if there are test failures, the jslint step will run after that and set the exit code to 0.

[rvagg]

@piscisaureus also vcbuild.bat release nosign x64 test-simple test-message only runs test-message and I don't understand why.

[piscisaureus]

@piscisaureus also vcbuild.bat release nosign x64 test-simple test-message only runs test-message and I don't understand why.

Huh, do you have a jenkins link? vcbuild.bat looks like it does the right thing.

[rvagg]

https://jenkins-iojs.nodesource.com/job/iojs+any-pr+multi/200/nodes=iojs-win2012r2/console for instance was run with that string. I'm also on one of the servers now running it myself and get the same behaviour. As if test-simple is completely ignored as an argument. Switch them around and it seems to work (although tbh I can't tell just at the moment if test-message is being run or not after test-simple).

[piscisaureus]

@rvagg oops. vcbuild.bat is the culprit:

https://github.com/iojs/io.js/blob/v1.x/vcbuild.bat#L59

[piscisaureus]

@rvagg #998