Test failure with upcoming mongodb extension 1.3.0

[remicollet]

Seems minor (connection fail with different exception)

There was 1 failure:

1) Alcaeus\MongoDbAdapter\Tests\Mongo\MongoCollectionTest::testFindOneConnectionIssue
Failed asserting that exception of type "MongoDB\Driver\Exception\InvalidArgumentException" matches expected exception "MongoConnectionException". Message was: "Failed to parse MongoDB URI: 'mongodb://localhost:28888?connectTimeoutMS=1'. Invalid host string in URI." at
/usr/share/php/MongoDB/Client.php:81
/builddir/build/BUILDROOT/php-alcaeus-mongo-php-adapter-1.1.2-1.fc26.remi.x86_64/usr/share/php/Alcaeus/Mongo/MongoClient.php:95
/builddir/build/BUILD/mongo-php-adapter-052cd5d1f0698b96993d8b1b613ff9c6d25b912f/tests/Alcaeus/MongoDbAdapter/TestCase.php:45
/builddir/build/BUILD/mongo-php-adapter-052cd5d1f0698b96993d8b1b613ff9c6d25b912f/tests/Alcaeus/MongoDbAdapter/Mongo/MongoCollectionTest.php:668
.

[remicollet]

@jmikola @derickr for your information

[remicollet]

Sorry, I forget the errors (seems important)

There were 4 errors:

1) Alcaeus\MongoDbAdapter\Tests\Mongo\MongoBinDataTest::testCreateWithBsonBinary
MongoDB\Driver\Exception\InvalidArgumentException: Expected UUID length to be 16 bytes, 3 given

/builddir/build/BUILD/mongo-php-adapter-052cd5d1f0698b96993d8b1b613ff9c6d25b912f/tests/Alcaeus/MongoDbAdapter/Mongo/MongoBinDataTest.php:42

2) Alcaeus\MongoDbAdapter\Tests\Mongo\MongoClientTest::testReadPreferenceOptionsAreInherited with data set "optionsArray" (array('secondaryPreferred', 'a:b'), 'mongodb://localhost', array(array('b')))
MongoDB\Driver\Exception\InvalidArgumentException: Expected array for "readPreferenceTags" URI option, string given

/usr/share/php/MongoDB/Client.php:81
/builddir/build/BUILDROOT/php-alcaeus-mongo-php-adapter-1.1.2-2.fc26.remi.x86_64/usr/share/php/Alcaeus/Mongo/MongoClient.php:95
/builddir/build/BUILD/mongo-php-adapter-052cd5d1f0698b96993d8b1b613ff9c6d25b912f/tests/Alcaeus/MongoDbAdapter/TestCase.php:45
/builddir/build/BUILD/mongo-php-adapter-052cd5d1f0698b96993d8b1b613ff9c6d25b912f/tests/Alcaeus/MongoDbAdapter/Mongo/MongoClientTest.php:126

3) Alcaeus\MongoDbAdapter\Tests\Mongo\MongoClientTest::testReadPreferenceOptionsAreInherited with data set "overridden" (array('secondaryPreferred', 'a:b'), 'mongodb://localhost/?readPref...gs=c:d', array(array('d'), array('b')))
MongoDB\Driver\Exception\InvalidArgumentException: Expected array for "readPreferenceTags" URI option, string given

/usr/share/php/MongoDB/Client.php:81
/builddir/build/BUILDROOT/php-alcaeus-mongo-php-adapter-1.1.2-2.fc26.remi.x86_64/usr/share/php/Alcaeus/Mongo/MongoClient.php:95
/builddir/build/BUILD/mongo-php-adapter-052cd5d1f0698b96993d8b1b613ff9c6d25b912f/tests/Alcaeus/MongoDbAdapter/TestCase.php:45
/builddir/build/BUILD/mongo-php-adapter-052cd5d1f0698b96993d8b1b613ff9c6d25b912f/tests/Alcaeus/MongoDbAdapter/Mongo/MongoClientTest.php:126

4) Alcaeus\MongoDbAdapter\Tests\Mongo\MongoClientTest::testReadPreferenceOptionsAreInherited with data set "multipleTagsetsOptions" (array('secondaryPreferred', 'a:b,c:d'), 'mongodb://localhost', array(array('b', 'd')))
MongoDB\Driver\Exception\InvalidArgumentException: Expected array for "readPreferenceTags" URI option, string given

/usr/share/php/MongoDB/Client.php:81
/builddir/build/BUILDROOT/php-alcaeus-mongo-php-adapter-1.1.2-2.fc26.remi.x86_64/usr/share/php/Alcaeus/Mongo/MongoClient.php:95
/builddir/build/BUILD/mongo-php-adapter-052cd5d1f0698b96993d8b1b613ff9c6d25b912f/tests/Alcaeus/MongoDbAdapter/TestCase.php:45
/builddir/build/BUILD/mongo-php-adapter-052cd5d1f0698b96993d8b1b613ff9c6d25b912f/tests/Alcaeus/MongoDbAdapter/Mongo/MongoClientTest.php:126

[alcaeus]

Thanks Remi! I'll add beta 1 to the build pipeline later today and get on fixing these issues! 👍

[jmikola]

@remicollet: Thanks for checking this as well!

1) Alcaeus\MongoDbAdapter\Tests\Mongo\MongoCollectionTest::testFindOneConnectionIssue
Failed asserting that exception of type "MongoDB\Driver\Exception\InvalidArgumentException" matches expected exception "MongoConnectionException". Message was: "Failed to parse MongoDB URI: 'mongodb://localhost:28888?connectTimeoutMS=1'. Invalid host string in URI." at

This appears to be due to a libmongoc change, as "/" is required to separate the host and query string. I'll ask @bjori and @ajdavis about that later, but I believe this behavior is consistent with the connection string spec, which states:

The host information section of the connection string is delimited by the trailing slash ("/") or end of string.

"?" is not considered to be a delimiter that ends host parsing.

1) Alcaeus\MongoDbAdapter\Tests\Mongo\MongoBinDataTest::testCreateWithBsonBinary
MongoDB\Driver\Exception\InvalidArgumentException: Expected UUID length to be 16 bytes, 3 given

We now validate argument length for Binary objects with UUID subtypes (PHPC-895).

2) Alcaeus\MongoDbAdapter\Tests\Mongo\MongoClientTest::testReadPreferenceOptionsAreInherited with data set "optionsArray" (array('secondaryPreferred', 'a:b'), 'mongodb://localhost', array(array('b')))
MongoDB\Driver\Exception\InvalidArgumentException: Expected array for "readPreferenceTags" URI option, string given

This and the related errors that follow are likely caused by throwing if a value in the URI options array has an unexpected type (PHPC-887). Previously, we would have simply ignored the option. I'm curious how the adapter was testing this, though. Were you purposefully constructing a client with invalid types and then checking that they were ignored, or did we just uncover a subtle gap in the adapter mimicking the ext-mongo API where it missed converting legacy option values to the new format?

[ajdavis]

That's right, I did some rewriting of the parser here:

https://jira.mongodb.org/browse/CDRIVER-2190

and in the process fixed a related bug, which is that we thought this was valid:

mongodb://host?option=value

It must be:

mongodb://host/?option=value

[alcaeus]

I'm curious how the adapter was testing this, though. Were you purposefully constructing a client with invalid types and then checking that they were ignored, or did we just uncover a subtle gap in the adapter mimicking the ext-mongo API where it missed converting legacy option values to the new format?

That last one would be my guess. Since the test passes against ext-mongo but no longer passes when converting for ext-mongodb, I'm guessing that the adapter will have to do some additional conversion there.