Release proposal: node-gyp@5.0.6

[rvagg]

Not sure how much this is needed but it might help npm, I don't see an indication which way they're going with node-gyp@6.

---

Commits on master that aren't on this branch:

• [68319a2c34] - v6.0.1: bump version and update changelog (Rod Vagg) #1935
• [1a4ff636d5] - v6.0.0: bump version and update changelog (Rod Vagg) #1892
• [dd0e97ef0b] - (SEMVER-MAJOR) lib: try to find python after python3 (Sam Roberts) #1907
• [f60ed47d14] - travis: add Python 3.5 and 3.6 tests on Linux (Christian Clauss) #1903
• [c763ca1838] - (SEMVER-MAJOR) doc: Declare that node-gyp is Python 3 compatible (cclauss) #1811
• [3d1c60ab81] - (SEMVER-MAJOR) lib: accept Python 3 by default (João Reis) #1844
• [c6e3b65a23] - (SEMVER-MAJOR) lib: raise the minimum Python version from 2.6 to 2.7 (cclauss) #1818

---

Commits on this branch that aren't on master:

• [d6041b0f1c] - v5.0.6: bump version and update changelog (Rod Vagg)
• [3891391746] - doc: reconcile README with Python 3 compat changes (Rod Vagg) #1911
• [07f81f1920] - lib: accept Python 3 after Python 2 (Sam Roberts) #1910
• [3834156a92] - test: add Python 3.5 and 3.6 tests on Linux (cclauss) #1909

---

(.travis.yml isn't materially different between the branches even with those commit discrepancies)

Changes in this release:

• [cdec00286f] - doc: adjustments to the README.md for new users (Dan Pike) #1919
• [b7c8233ef2] - test: fix Python unittests (cclauss) #1961
• [e12b00ab0a] - doc: macOS Catalina add two commands (Christian Clauss) #1962
• [70b9890c0d] - test: add header download test (Rod Vagg) #1796
• [4029fa8629] - test: configure proper devDir for invoking configure() (Rod Vagg) #1796
• [fe8b02cc8b] - doc: add macOS_Catalina.md document (cclauss) #1940
• [8ea47ce365] - gyp: python3 fixes: utf8 decode, use of 'None' in eval (Wilfried Goesgens) #1925
• [c7229716ba] - gyp: iteritems() -> items() in compile_commands_json.py (cclauss) #1947
• [2a18b2a0f8] - gyp: make cmake python3 compatible (gengjiawen) #1944
• [70f391e844] - gyp: fix TypeError in XcodeVersion() (Christian Clauss) #1939
• [9f4f0fa34e] - gyp: finish decode stdout on Python 3 (Christian Clauss) #1937
• [7cf507906d] - src,win: allow 403 errors for arm64 node.lib (Richard Lau) #1934
• [ad0d182c01] - deps: update deps to roughly match current npm@6 (Rod Vagg) #1920
• [1553081ed6] - test: upgrade Linux Travis CI to Python 3.8 (Christian Clauss) #1923
• [0705cae9aa] - travis: ignore failed brew upgrade npm, update xcode (Christian Clauss) #1932
• [7bfdb6f5bf] - build: avoid bare exceptions in xcode_emulation.py (Christian Clauss) #1932
• [7edf7658fa] - lib,install: always download SHA sums on Windows (Sam Hughes) #1926
• [69056d04fe] - travis: add Windows + Python 3.8 to the mix (Rod Vagg) #1921

---

The main risk here is in the dependency version changes from #1920.

[cclauss]

Nice work as always @rvagg

We need to get Travis to be green again and also, given nodejs/node-v8#119 (comment) perhaps we should set our lower limit to be Python >= 2.7.9.

[rvagg]

Sorry, you'll have to spoon feed me, i don't grok what's going on - does this demonstrably break for users of 2.6 (if there are any) or is this a linting failure or test-specific failure?
Bumping minimum Python version should be considered semver-major as we've discussed before so if we can just maintain where we're at with 5.x then that would be best. Is there a commit I can roll back here?

[cclauss]

The change in question is not in this repo but is in our dependency nodejs/node-v8#119 (comment). That code makes it convoluted to support both Python < 2.7.9 and Python 3.

[rvagg]

it looks like npm wants to stick with v5 for npm v6, so we should still maintain these

I've published 5.0.6 from this, but we have a few things in the queue to warrant a 5.1.0 soon:

• [3bcba2a01a] - (SEMVER-MINOR) lib: noproxy support, match proxy detection to request (Matias Lopez) #1978
• [345c70e56d] - test: direct python invocation & simpler pyenv (Matias Lopez) #1979
• [d6a7e0e1fb] - test: fix macOS Travis on Python 2.7 & 3.7 (Christian Clauss) #1979
• [5a64e9bd32] - test: initial Github Actions with Ubuntu & macOS (Christian Clauss) #1985
• [f7b6b6b77b] - doc: fix typo in README.md (Fix typo in README.md #1985) (Suraneti Rodsuwan) #1985
• [6b8f2652dd] - doc: add travis badge (Rod Vagg) #1971

I'd like to see that proxy change bake in the wild first, I'm sure there are people that will grab v6 just for that one.

[rvagg]

I had to publish a v5.0.7 on top of this, same commits but with a pack file removed from the pack file ... something i've done far too often but keep on doing. I ran npm pack to inspect the contents of what's being published, but then didn't remove the pack file so an npm publish packs up the pack file.

[richardlau]

I had to publish a v5.0.7 on top of this, same commits but with a pack file removed from the pack file ... something i've done far too often but keep on doing. I ran npm pack to inspect the contents of what's being published, but then didn't remove the pack file so an npm publish packs up the pack file.

That's what --dry-run is for? I'm not opposed to adding the pack file to .gitignore/.npmignore or explicitly listing the files to include in package.json (like we did in CITGM: nodejs/citgm#704).

[rvagg]

That's what --dry-run is for?

yeah, probably, it's just a habit I need to break at some point, my fault and I don't think we need special handling for it