build: sort .PHONY rules and add test-gc-clean

[joyeecheung]

The first commit sorts the .PHONY rules for the ease of updating and backporting changes to this list.
The second commit adds a test-gc-clean rule to clean the files generated during make test-gc

Checklist

• make -j4 test (UNIX), or vcbuild test (Windows) passes
• commit message follows commit guidelines

Affected core subsystem(s)

build

[joyeecheung]

CI: https://ci.nodejs.org/job/node-test-pull-request/7053/

[Trott]

@nodejs/build

[joyeecheung]

CI errors don't seem to be related..pinging @nodejs/build again

[Trott]

LGTM

[gibfahn]

Should we be using test-gc-clean elsewhere in the Makefile (e.g. in clean)?

[joyeecheung]

@gibfahn I am still trying to put together a make test-clean(or, an additional test-clean-all for tests that don't get run in a simple make test)..I think make clean should be the reverse of make, not make test.

[gibfahn]

@joyeecheung I think make clean should clean up everything and take you back to a pristine state (as if I'd cloned a fresh copy of the repo). At least for me, the aim is to have the Makefile be as intuitive as possible, so contributors have to delve into its rules as little as possible.

If you say "I want to clean up", you probably want to clean up everything (make clean -> a clean directory). If you say "I only want to clean up X", then you'd probably expect a make clean-X. The latter case is probably rare, how often do you only want to clean up one thing?

Cleaning up more than you expected is annoying (you have to rebuild everything). Cleaning up less than you expected can be confusing (why are my gc tests still passing even though I actually broke native modules).

Also, make is a shortcut for make all, so to be consistent you'd have to have a make clean-all, which would be way too confusing (it would only clean up build artefacts).

---

If you wanted to add a make build-clean or something then I guess that might be useful, although if you're wiping the build you almost certainly want to regenerate the test artifacts.

I see the reason for having a make test-clean, but not for having a separate make test-clean-all. In what scenario do you not want to clean up all test artifacts? I don't think there's much overhead to always trying to clean gc (for example) even if it's not there.

[joyeecheung]

@gibfahn Yeah I am convinced make clean should bring the repo back to a clean slate now.

As for make test-clean-all, that's part of the "bring it back to a clean slate" thing, since the artifact generations are scattered around in different rules, I think it would be nice to have a rule showing all the possible artifacts / non-standard tests. It's not really meant to be run directly...

[gibfahn]

As for make test-clean-all, that's part of the "bring it back to a clean slate" thing, since the artifact generations are scattered around in different rules, I think it would be nice to have a rule showing all the possible artifacts / non-standard tests. It's not really meant to be run directly...

In my mind you'd have:

clean: clean-test clean-build [...?] clean-test: clean-test-gc clean-test-addons [...] 

So clean-test would do what I think you mean by clean-test-all. You might have a clean-test-fixtures or something, which would just clean up the artifacts from the default test suites.

[joyeecheung]

@gibfahn What I had in mind is:

clean: test-clean-all clean-build ... test-clean-all: test-clean test-addons-clean test-gc-clean test-clean: # clean up stuff generated during make test 

Anyways, one more CI for this PR...https://ci.nodejs.org/job/node-test-pull-request/7164/

[jasnell]

This needs a rebase before it can land

[joyeecheung]

Rebased. New CI: https://ci.nodejs.org/job/node-test-pull-request/7278/

[joyeecheung]

CI failures look unrelated. Landed in 9decfb1...baa2602, thanks!

[evanlucas]

This is not landing cleanly on v7.x-staging. Mind submitting a backport?

[gibfahn]

Should this be backported to v6.x-staging? If yes please follow the guide and raise a backport PR, if no let me know or add the dont-land-on label.