brew tests: filter out vendor/

[mistydemeo]

• [x ] Have you followed the guidelines in our Contributing document?
• Have you checked to ensure there aren't other open Pull Requests for the same change?
• Have you added an explanation of what your changes do and why you'd like us to include them?
• Have you successfully run brew tests with your changes locally?

---

Not sure why this doesn't consistently repro for me, but I noticed that brew tests was failing for me because the test glob was picking up test files from the bundled gems installed in test/vendor/bundle.

[reitermarkus]

Remove test/vendor/bundle and test/.bundle. The bundled Gems have been in vendor/bundle for a few months now, but it seems that the test/.bundle/config still contains the BUNDLE_PATH for some people.

We could add test/.bundle/config to version control, to make sure it contains the same for everyone.

[mistydemeo]

Adding it to version control or explicitly running bundle config to set the path makes sense to me. But I still think we should exclude these files, because even if the vendor bundle gets installed in the right place, this will fail if a user happens to have this old directory left around.

[MikeMcQuaid]

We could add test/.bundle/config to version control, to make sure it contains the same for everyone.

Makes sense to me 👍

[MikeMcQuaid]

@DomT4 Does this fix #2184 for you?

[DomT4]

@MikeMcQuaid No, same issue. To be honest I initially pinned this yesterday on an upstream problem at some level, because json have had identical issues reported to them in the past with semi-new Ruby x.Y.x releases. I noted it upstream on the long-running issue but nobody else has reported it as far as I can see, so I'm not sure what's going on entirely.

[DomT4]

If I gem install json -V outside of Bundle masking the output there's a noted failure on an element, but it's not the element I was seeing blow up tests.

linking shared-object json/ext/parser.bundle
current directory: /usr/local/var/rbenv/versions/2.4.0/lib/ruby/gems/2.4.0/gems/json-2.0.3/ext/json/ext/parser
make "DESTDIR=" install
/usr/local/bin/ginstall -c -m 0755 parser.bundle ./.gem.20170227-65081-2nbc27/json/ext
To see why this extension failed to compile, please check the mkmf.log which can be found here:

  /usr/local/var/rbenv/versions/2.4.0/lib/ruby/gems/2.4.0/extensions/x86_64-darwin-16/2.4.0-static/json-2.0.3/mkmf.log

...

"clang -o conftest -I/usr/local/var/rbenv/versions/2.4.0/include/ruby-2.4.0/x86_64-darwin16 -I/usr/local/var/rbenv/versions/2.4.0/include/ruby-2.4.0/ruby/backward -I/usr/local/var/rbenv/versions/2.4.0/include/ruby-2.4.0 -I. -I/usr/local/var/rbenv/versions/2.4.0/include  -D_XOPEN_SOURCE -D_DARWIN_C_SOURCE -D_DARWIN_UNLIMITED_SELECT -D_REENTRANT    -O3 -Wno-error=shorten-64-to-32  -pipe conftest.c  -L. -L/usr/local/var/rbenv/versions/2.4.0/lib -L. -L/usr/local/var/rbenv/versions/2.4.0/lib  -fstack-protector -L/usr/local/lib     -lruby-static -framework CoreFoundation  -lpthread -lgmp -ldl -lobjc "
conftest.c:15:57: error: use of undeclared identifier 'rb_enc_raise'
int t(void) { void ((*volatile p)()); p = (void ((*)()))rb_enc_raise; return !p; }
                                                        ^
1 error generated.

[ilovezfs]

@DomT4 does it happen if you checkout a tag? 1.1.10, 1.1.9, 1.1.8, ... wondering if this is a bisectable problem.

[DomT4]

json 1.8.6 builds without errors on the gem install at least. None of 2.0.0, 2.0.1, 2.0.2 or the latest 2.0.3 compile cleanly. On the brew side, no error thrown at 9cd5a21, 619791e, 9cce341, e59ada5, 664d0c6.

Had appeared by the time 69d1ced was cut.

[ilovezfs]

OK. So you should be able to do

git checkout master
git bisect start
git checkout 664d0c6
git bisect good
git checkout 69d1ced
git bisect bad

And then tell it git bisect good or git bisect bad for each commit it asks you to test until it spits out the first bad commit.

[DomT4]

Was already working on it 😉. e40c63f is where it was introduced, merged as part of #1913.

[ilovezfs]

🎉

[reitermarkus]

Make this test/vendor/bundle, since there is at least one test actually testing vendored code (plist, currently still in cask-tests).

[mistydemeo]

Done!

[DomT4]

Was it intentional here that the repo becomes "dirty" after brew tests is run?

cd $(brew --repo)
git fetch && git reset --hard origin/master
brew tests
brew --version
Homebrew 1.1.10-432-ge02223960-dirty

The change is:

diff --git a/Library/Homebrew/test/.bundle/config b/Library/Homebrew/test/.bundle/config
index dcf3d1c4c..e451829e9 100644
--- a/Library/Homebrew/test/.bundle/config
+++ b/Library/Homebrew/test/.bundle/config
@@ -1,3 +1,3 @@
 ---
 BUNDLE_PATH: "../vendor/bundle"
-BUNDLE_DISABLE_SHARED_GEMS: '1'
+BUNDLE_DISABLE_SHARED_GEMS: "true"