Remove single install

[gabyx]

Thats the changeset without any symlinks. Tests seem to work so far.

We need to check if we made all path from older update clear to fail. I hope we did. Here nothing really much changed.

I refactored also in this PR

• moved githooks.autoupdate.registered -> githooks.registered and moved the file $INSTALL_DIR/autoupdate/registered to $INSTALL_DIR/registered (maybe other name?) legacy_transformtion_end deletes the whole $INSTALL_DIR/autoupdate/ folder because that was the only need for it...

Note: its good to do this right here, because the registered list only contained non-single installs. And I think the feature to have now all repos registered isnt that bad since, we could potentially act on this list in future updates. I would this feature still in. might be handy ...

[gabyx]

Shall I also move the autoupdate changes out of this PR. Its a bit more, work, hm... Are you able to look over all this renaming, would be greate, otherwise I try to remove it and add it back in in another PR which succeeds this one...? let me know

[coveralls]

Pull Request Test Coverage Report for Build 704

• 53 of 57 (92.98%) changed or added relevant lines in 3 files are covered.
• 2 unchanged lines in 2 files lost coverage.
• Overall coverage decreased (-0.4%) to 80.93%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
install.sh	44	46	95.65%
uninstall.sh	4	6	66.67%
<!--	Total:	53	57	92.98%	-->

Files with Coverage Reduction	New Missed Lines	%
install.sh	1	79.45%
uninstall.sh	1	84.05%
<!--	Total:	2		-->

Totals
Change from base Build 699:	-0.4%
Covered Lines:	2071
Relevant Lines:	2559

---

💛 - Coveralls

[rycus86]

There are some test failures here, but otherwise the change looks OK, you can leave the config move in this PR if you want.

One question I have is whether we need the legacy transformations in the uninstall script as well? Don't we fetch that with curl? If yes, we'd potentially get a newer version than what you have locally, right?

[gabyx]

nono :-) we never use curl anymore, everything goes over the release clone :-) the legacy stuff should be in the uninstall script as well to prevent dangling old stuff...

[gabyx]

I look at the tests later thanks!

[rycus86]

nono :-) we never use curl anymore, everything goes over the release clone :-)

Hm, https://github.com/rycus86/githooks/blob/master/README.md#uninstalling uses curl to fetch the uninstall script, does that internally update the release clone now to just delete it?

[gabyx]

nono :-) we never use curl anymore, everything goes over the release clone :-)

Hm, https://github.com/rycus86/githooks/blob/master/README.md#uninstalling uses curl to fetch the uninstall script, does that internally update the release clone now to just delete it?

Ah hm.. we might have some inconsistency here, you are right. One should really call git hooks uninstall and not use the curl method preferably? But we can also check in uninstall.sh if there is a release clone, if true we dispatch to the install script there... other wise we just continue with the script...?

[rycus86]

nono :-) we never use curl anymore, everything goes over the release clone :-)

Hm, https://github.com/rycus86/githooks/blob/master/README.md#uninstalling uses curl to fetch the uninstall script, does that internally update the release clone now to just delete it?

Ah hm.. we might have some inconsistency here, you are right. One should really call git hooks uninstall and not use the curl method preferably? But we can also check in uninstall.sh if there is a release clone, if true we dispatch to the install script there... other wise we just continue with the script...?

Yeah, I'm not too bothered how it happens tbh as long as it works for old and new installations I guess on a best effort.

[gabyx]

I will take a look at this a new PR. Pushed again, hopefully the tests now work. Forgot to call prepare_target_template_dir always... Good we have tests :-)

[rycus86]

Looks better, still some core.hooksPath issues it seems?

[gabyx]

Jeah test-067.sh was the problem. I thought I can remove

# Using core.hooksPath implies it applies to all repo's
    if is_single_repo_install && use_core_hookspath; then
        echo "! Cannot use --single and --use-core-hookspath together" >&2
        return 1
    fi

which was wrong...

[gabyx]

Question: Shouldn't all .sh scripts inside tests be executable. Why is no .sh script executable in the repo?

[rycus86]

Question: Shouldn't all .sh scripts inside tests be executable. Why is no .sh script executable in the repo?

I always call them with sh <something>.sh so I never bothered to set them I think. The actual steps should also not run by themselves, only through the test runner I guess.

Is it causing any problems for you or just curious?

[gabyx]

Nono, I was just thinking, ah ok, tough the reasoning is all scripts which should not be executable by them self shouldn't be executable.

Maybe make all tests-*.sh scripts executable. though I dont have to write all ways sh ... :)?

[rycus86]

Yeah, put up a PR if you have strong feelings about those test runners. :)

[rycus86]

Looks like Travis is happy, let me try to review this then I'll merge it. 🎉

[gabyx]

Thx!

[rycus86]

OK, nothing serious, added a few comments for you to look at, please just open another PR if you think there's something to do with them. Merging this one now.