Example from issue:
```php
<?php
#
# Comment
#
$a = 100;
```
The problem with the example from the issue is that `#` is immediately followed by a line break.
And since the rule in the lexer for such comments was changed, this case was handled incorrectly.
```
(('#' ^'[') | '//') any_line* when is_not_comment_end => {
lex.ungetStr("?>")
lex.addFreeFloatingToken(tkn, token.T_COMMENT, lex.ts, lex.te)
};
```
This rule has one problem, it checks two characters at once, first for the match `#`, and
then for the mismatch `[`, which leads to the fact that in the case of an empty comment, the first
matcher will capture `#`, and the second line break (`\n`), which will lead to the fact that `any_line`
matcher will not work and will not increase the line number.
The next rule added is specifically for this case.
```
'#' newline when is_not_comment_end => {
lex.ungetStr("?>")
lex.addFreeFloatingToken(tkn, token.T_COMMENT, lex.ts, lex.te)
};
```
Renamed 'possible_comma' with 'optional_comma' as in PHP-Parser
Replace all 'identifier' with 'identifier_ex' and all 'T_STRING' with 'identifier' as in PHP-Parser
PHP 8 Update
- nullsafe operator (?->)
- Remove (real) cast
- Named arguments
- Remove (unset) cast
- Remove {} access
- match expression
- Union types in type hints and static typehint
- Block catch without variable
- Trailing comma in parameter lists
- throw can be used as an expression
- Concatenation precedence
- Declaring properties in the constructor
- Attributes
- Names in the namespace are treated as a single token
- Trailing comma in closure use list
- Check that ::class on object works
- Deferencable changes and arbitrary expressions in new/instanceof
1. Now, for the expression 'new A' the correct
values EndLine, EndPos, and not -1 will be set;
2. Also, for expressions from php5 '$a = &new Foo',
the condition for parsing is fixed when it is necessary
to set the Args values and the initialization of the
NewTkn field is added, in the case when this condition
is false.
## Problem description
The reason why the positions after parsing became
incorrect is that the check that is responsible for
separating expressions like 'new A' and 'new A (args)'
relied on comparison with nil, however, when the parser
was updated, 'ctor_arguments' began to return not nil,
but &ArgumentList{}, so the condition was always true,
and in this case, when calculating the position, the
second argument of the 'NewTokenNodePosition' function
was nil, which is why -1 was returned there.
For the second, the reasons are similar. In addition,
there was a mistake in the number that needs to be
checked. In the expression:
'variable' '=' '&' T_NEW class_name_reference ctor_arguments'
it is necessary to check not 3, but 6 elements for nil.
Laytan Laats
i582
Tsygankov-Slava
Ruud Kamphuis
i582
Vadym Slizov