PHPCSStandards / PHP_CodeSniffer / 19056334694 / 8
79%
master: 79%

LAST BUILD BRANCH: feature/fixer-conflict/PSR12.Functions.ReturnTypeDeclaration
DEFAULT BRANCH: master
Ran
Files 310
Run time 10s
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
coverage: 0.894% (-0.001%) from 0.895%
19056334694.8

push

github

web-flow 
PHP 8.5 | Tokenizer/PHP: properly fix "Using null as an array offset" deprecation (#1230)

If a slash/hash comment is on its own line, the new line token is merged into the comment token for PHP 8.0+, and the new line is skipped by setting it to `null`.

The PHP Tokenizer did not take the possibility of a token being set to `null` into account when retokening a `?` to either `T_NULLABLE` or `T_INLINE_THEN`, leading to the PHP 8.5 deprecation notice.
In that case, the sniff first looks forward to see if we can draw a conclusion based on the non-empty tokens following the `?` and if not, walks backward to look at the tokens before the `?`.
The problem occurs when a `null` token exists in the sequence before the `?`.

This `null` token will only have been injected with a specific code sequence: when there is a slash/hash comment followed by a new line in the token sequence before the `?` and there is no indentation/new line whitespace on the next line.

Also see the detailed analysis by andrewnicols in PHPCSStandards/PHP_CodeSniffer 1216#issuecomment-3274198443

There are only two situations in which this causes the tokenizer to hit the PHP 8.5 "Using null as an array offset" deprecation notice:
1. If the `?` token is the last token in the file (live coding - the issue was discovered via a test related to live coding).
2. If there are tokens after the `?`, but not tokens which allows us to draw a conclusion (yet) and there is a slash/hash comment between the `?` and the previous non-empty token.

This commit:
1. Adds dedicated tests for both situations.
2. Adds a new fix for situation [1] as if there are no tokens after a `?`, we cannot determine definitively whether the `?` is a nullable operator or an inline then. For BC, this should be tokenized as `T_INLINE_THEN`.
3. Makes a small tweak to the previously added fix which addressed situation [2].
    Alternatively, we could consider switching to using the "$finalTokens" token stack to do the token walk... (continued)

313 of 34995 relevant lines covered (0.89%)

0.02 hits per line

Source Files on job cbf-os-ubuntu-latest-ubuntu-latest-php-7.2-custom-ini-true - 19056334694.8