This fix allows for more granular address control when marshalling nodes
across WASM. This is necessary for node methods which accept another
node as a parameter (i.e., `childWithDescendant()`)
**Problem:** After `ts_parser_parse_with_options()`, the parser options
are still stored in the parser object, meaning that a successive call to
`ts_parser_parse()` will actually behave like
`ts_parser_parse_with_options()`, which is not obvious and can have
unintended consequences.
**Solution:** Reset to empty options state after
`ts_parser_parse_with_options()`.
* Rename corpus test functions to allow easy filtering by language
* Use usize for seed argument
* Avoid retaining useless stack versions when reductions merge
We found this problem when debugging an infinite loop that happened
during error recovery when using the Zig grammar. The large number of
unnecessary paused stack versions were preventing the correct recovery
strategy from being tried.
* Fix leaked lookahead token when reduction results in a merged stack
* Enable running PHP tests in CI
* Fix possible infinite loop during error recovery at EOF
* Account for external scanner state changes when detecting changed ranges in subtrees
When using TypeScript projects using other module settings than CommonJs, the types were not correctly exposed, and the compilation failed.
This adds the types path to the exports so compilation works for `module: NodeNext` and other variants.
**Problem:** When resetting the parser during subtree balancing, an
error is thrown:
```
parser.c:2198: ts_parser_parse: Assertion `self->finished_tree.ptr' failed.
```
**Solution:** Reset `canceled_balancing` to false in
`ts_parser_reset()`.
Problem: Macros (re)defined in `endian.h` conflict with system headers
on FreeBSD (at least).
Solution: Rely on system `endian.h` on OpenBSD, FreeBSD, NetBSD, and
DragonFly
Ref. https://github.com/mikepb/endian.h/issues/4
This allows users to bail parsing if an error was *definitely* detected
using the progress callback, as all possible stack versions have a
non-zero error cost.
Co-authored-by: Amaan Qureshi <amaanq12@gmail.com>
