* Run tests in random order
Being able to run tests in any order is a pre-requisite for being able
to run them in parallel.
* Reset type coverage between tests, fix affected tests
* Reset parser and lexer between test runs and on php version change
Previously lexer was reset, but parser kept the reference to the old
one, and reference to the parser was kept by StatementsProvider. This
resulted in order-dependent tests - if the parser was first initialized
with phpVersion set to 7.4 then arrow functions worked fine, but were
failing when the parser was initially constructed with settings for 7.3
This can be demonstrated on current master by upgrading to
nikic/php-parser:4.9 and running:
```
vendor/bin/phpunit --no-coverage --filter="inferredArgArrowFunction" tests/ClosureTest.php
```
Now all tests using PHP 7.4 features must set the PHP version
accordingly.
* Marked more tests using 7.4 syntax
* Reset newline-between-annotation flag between tests
* Resolve real paths before passing them to checkPaths
When checkPaths is called from psalm.php the paths are resolved, so we
just mimicking SUT behaviour here.
* Restore newline-between-annotations in DocCommentTest
* Tweak Appveyor caches
* Tweak TravisCI caches
* Tweak CircleCI caches
* Run tests in parallel
Use `vendor/bin/paratest` instead of `vendor/bin/phpunit`
* Use default paratest runner on Windows
WrapperRunner is not supported on Windows.
* TRAVIS_TAG could be empty
* Restore appveyor conditional caching
I noticed that entries in InternalTaintSinkMap don't warn
unless the functions are added to CallMap.php
e.g. `pg_exec($conn, $_GET['query']);`
(pg_exec is an alias of pg_query)
`class_exists()` interacts with PHP's autoloader feature, which allows
user-defined behaviour to take place when PHP tries to load a given
class or interface.
* add test case for DeprecatedClass suppress on property
* fix test case for DeprecatedClass suppress on property - suppress MissingConstructor also
* ugly way to fix - double analyze propery statement to get dockblock + fix in TypeChecker - respect given suppressed_issues in checkNamedObject
* replace double analyze propery with direct DocBlock manipulation
* replace double analyze propery with direct DocBlock manipulation
* add exception handler to keep original addContextProperties behavior
* correct way to unwrap value from array
This change does a few things:
* Renames `BeforeAnalyzeFileInterface` to `BeforeFileAnalysisInterface`.
* Adds a few more parameters to `BeforeFileAnalysisInterface`.
* Since the analysis was never called due to a bug, the above two
changes are safe to do.
* Fix the above-mentioned bug, and now adds the hooks to the config.
* Update the documentation about the File-related analysis plugin hooks.
* Add a test to ensure that this does not break again in the future.
Fixes: #3969
* narrowed `reset` return type
BUT psalm seems to ignore the stub
* narrowed `reset` and `end` return type, this time for real
* fixed UnusedVariable Issue
* fixed RedundantCondition Issue
caused by `end`s return type being more precise
* Improve solution slightly
Co-authored-by: Matthew Brown <github@muglug.com>
* Use more consistent language for Internal errors
There's no meaningful distinction between something being internal and
being 'marked as internal'.
* Rename property psalm_internal to internal in all storage classes
This property holds metadata that can be set using either @internal
or @psalm-internal in docblocks
* Change types of internal properties in storage from ?string to string
Simpler type is easier to handle. Non-internal methods can be considered
to be internal to the entire universe of PHP code, i.e. that code whose
namespace starts with the empty string. It's not a special case.
* Add passing tests for property fetch on an @internal class
I'm trying to work out why the equivilent InvalidCodeParse test is
failing for PsalmInternal
* Treat all properties of a psalm-internal class as psalm-internal
* Remove all $internal properties from storage - use psalm_internal instead
@internal can be represented as internal to the namespace root, avoiding
the need to check for both properties in storage later.
* Raise InternalClass issue when an internal class is used with e.g. instanceOf
* fix docs and tests
* Add return type declartion to code example in doc
* Don't allow class psalm-internal to overide a tighter method psalm-internal
* Break up long line
* Code style - move && from EOL to SOL
* Restore misplaced &&
* Fix code style
* Fix namespace fetching so it works
Co-authored-by: Matthew Brown <github@muglug.com>
* Fix#3810 removing ignore-nullable-return form stubs of key, array_key_first, array_key_last
* fixed test failing due to changes to key()
* Improve key return type
* Remove unnecessary check
Co-authored-by: Matthew Brown <github@muglug.com>
* Fix going to definition on return type
If a return type of a method or function is set incorrectly (with the PHP doc), then the references are not added for the `function() : MyClass` symbol, so the "Go to definition" feature of the LSP won't work. I don't believe an invalid return type or not should stop the symbol location being tracked (and not allowing code navigation).
In moved the symbol location tracking to be before the return early short circuit.
* Update SymbolLookupTest.php
Co-authored-by: Matthew Brown <github@muglug.com>
Currently it's not possible to "Go to definition" (LSP) on nullable args like `function( ?MyClass )` as the reference is stored a `MyClass|null` in the reference map, which will now resolve to a class name.
This PR removed any nullable type from the union before adding it to the reference map (as the reference map is only use to indicate a symbol was used in a given location, I think this makes sense).
In many sitations, stub files will receive a shallow _and_ deep scan when project files require extra analysys on things like parent classes. This makes stub file scanning inconsistent (orders become much less predictable for example), and adds extra process time to scan the files twice. In the case of stubs providing classes and functions for large projects, this is a non-trivial amount of time.
As deep scanning stubs should take just about as long as a shallow scan, it makes sense to just always deep scan them.
Fixes#3568.
This is useful for use cases such as saving multiline taint detection results.
Only the compact and console reports seem to use color right now.
In many cases, adding color codes to a text file would make it harder to read
in an editor.
E.g. `psalm --report=report.txt --report=report.pylint`.
It may be useful to have both a machine-readable and human readable
report, e.g. during taint analysis.
Previously, psalm would generate neither report if the report option was
repeated.
