[Validator] Fix call to undefined getParser() in YamlValidator

[yoeunes]

Q	A
Branch?	7.3
Bug fix?	yes
New feature?	no
Deprecations?	no
Issues	n/a
License	MIT

The set_error_handler closure was calling the non-existent method getParser(), which would cause a fatal error if a deprecation were ever triggered during parsing.

The fix instantiates the Parser as a local variable and passes it to the closure, resolving the issue and ensuring the validator remains stateless.

[stof]

could you add a test covering this behavior to avoid regressions ?

[yoeunes]

Hi @stof, I've added a new test that simulates deprecations on various lines, which tests that they are all correctly caught and converted into a ParseException.

[stof]

instead of introducing a getParser protected method in the constraint validator just for testing purpose (but which will then have to be covered by BC as it is not even internal), I would prefer to trigger a deprecation during an actual parsing to allowing keeping the parser fully local to the validator.
The lazy way would be to rely on one of the deprecations of the Yaml parser itself. But this won't be future-proof as those deprecations will become errors in 8.0. However, I have an idea that should work (I haven't tested it to confirm it):

1. Create a fixture object in the testsuite that triggers a deprecation in its __unserialize method
2. Create a Yaml string using !php/object with this object
3. Run the validator with a constraint providing the Yaml::PARSE_OBJECT flag

[yoeunes]

Hi @stof, you were right about the BC break, so I've implemented the test using your __unserialize idea instead and it works perfectly

[nicolas-grekas]

Thank you @yoeunes.