Lines with just spaces are being prepended to the next line

[pento]

Given the following test case (note the \t on the blank line between $foo and $bar):

<?php

$foo = 1;
    
$bar = 2;

The \t is being prepended to the $bar line.

This isn't a core bug, just something I noticed while testing. 🙂

Tested with PHPCS master, WPCS develop.

[jrfnl]

Hi @pento I've tried to reproduce the effect, but no luck. When I run the fixer against this snippet, I see the tab (correctly) being trimmed by the Generic.WhiteSpace.SuperfluousWhiteSpace sniff.

Got any more info I can use to try and reproduce this ?

[pento]

Not sure what I'm doing wrong, my testing setup is a copy of https://develop.svn.wordpress.org/trunk, this phpcs.xml.dist, and a composer.json like so:

{
	"require-dev": {
		"dealerdirect/phpcodesniffer-composer-installer": "*",
		"wp-coding-standards/wpcs": "dev-develop",
		"squizlabs/php_codesniffer": "dev-master"
	}
}

Running vendor/bin/phpcbf foo.php with foo.php containing the test case, I get the incorrect result.

[jrfnl]

My local dev setup is of course slightly different. I've got a phpcs.xml.dist which is a little ahead of what you're using, but nothing significant for this issue. I've just uploaded it to the ticket.

Other than that I have a branch in both PHPCS as well as WPCS which I keep up to date with the latest patches submitted related to the Core project.

These are the current versions (remind me to push/update them if they seem out of date):

• For WPCS, this is currently the same as the Core: Add Generic.Functions.FunctionCallArgumentSpacing sniff #1193 branch.
  https://github.com/jrfnl/PHP_CodeSniffer/tree/WP-Core/test-run
• For PHPCS, there are two open PRs merged into that branch - Generic/DisallowSpaceIndent: multi line comments and the case of the moving space squizlabs/PHP_CodeSniffer#1599, Squiz.Commenting.InlineComment will fail to fix comments at the end of the file squizlabs/PHP_CodeSniffer#1645. Additionally, Squiz/SemicolonSpacing: improve handling of semicolons in for control structures squizlabs/PHP_CodeSniffer#1691 would be a nice to have as well.
  https://github.com/jrfnl/WordPress-Coding-Standards/tree/WPCore-test-run

None of these should explain the difference though....

[jrfnl]

@pento Just thinking - could you run phpcbf over the test file with -vv and post the bottom bit of the output you receive ? Or even the complete output ? That should help me figure out what is going on.

[pento]

Soooo..... doing that made me realise it wasn't really a \t, it was actually four spaces. 🙂

Here's the `phpcbf` output

Processing foo.php 
	*** START PHP TOKENIZING ***
	Process token [0]: T_OPEN_TAG => <?php\n
	Process token [1]: T_WHITESPACE => \n
	Process token [2]: T_VARIABLE => $foo
	Process token [3]: T_WHITESPACE => ·
	Process token  4 : T_EQUAL => =
	Process token [5]: T_WHITESPACE => ·
	Process token [6]: T_LNUMBER => 1
	Process token  7 : T_SEMICOLON => ;
	Process token [8]: T_WHITESPACE => \n\t\n
	Process token [9]: T_VARIABLE => $bar
	Process token [10]: T_WHITESPACE => ·
	Process token  11 : T_EQUAL => =
	Process token [12]: T_WHITESPACE => ·
	Process token [13]: T_LNUMBER => 2
	Process token  14 : T_SEMICOLON => ;
	Process token [15]: T_WHITESPACE => \n
	*** END PHP TOKENIZING ***
	*** START TOKEN MAP ***
	*** END TOKEN MAP ***
	*** START SCOPE MAP ***
	*** END SCOPE MAP ***
	*** START LEVEL MAP ***
	Process token 0 on line 1 [col:1;len:5;lvl:0;]: T_OPEN_TAG => <?php\n
	Process token 1 on line 2 [col:1;len:0;lvl:0;]: T_WHITESPACE => \n
	Process token 2 on line 3 [col:1;len:4;lvl:0;]: T_VARIABLE => $foo
	Process token 3 on line 3 [col:5;len:1;lvl:0;]: T_WHITESPACE => ·
	Process token 4 on line 3 [col:6;len:1;lvl:0;]: T_EQUAL => =
	Process token 5 on line 3 [col:7;len:1;lvl:0;]: T_WHITESPACE => ·
	Process token 6 on line 3 [col:8;len:1;lvl:0;]: T_LNUMBER => 1
	Process token 7 on line 3 [col:9;len:1;lvl:0;]: T_SEMICOLON => ;
	Process token 8 on line 3 [col:10;len:0;lvl:0;]: T_WHITESPACE => \n
	Process token 9 on line 4 [col:1;len:4;lvl:0;]: T_WHITESPACE => ····\n
	Process token 10 on line 5 [col:1;len:4;lvl:0;]: T_VARIABLE => $bar
	Process token 11 on line 5 [col:5;len:1;lvl:0;]: T_WHITESPACE => ·
	Process token 12 on line 5 [col:6;len:1;lvl:0;]: T_EQUAL => =
	Process token 13 on line 5 [col:7;len:1;lvl:0;]: T_WHITESPACE => ·
	Process token 14 on line 5 [col:8;len:1;lvl:0;]: T_LNUMBER => 2
	Process token 15 on line 5 [col:9;len:1;lvl:0;]: T_SEMICOLON => ;
	Process token 16 on line 5 [col:10;len:0;lvl:0;]: T_WHITESPACE => \n
	*** END LEVEL MAP ***
	*** START ADDITIONAL PHP PROCESSING ***
	*** END ADDITIONAL PHP PROCESSING ***
[PHP => 17 tokens in 5 lines]... 
DONE in 24ms (1 fixable violations)
	=> Fixing file: 1/1 violations remaining	PHP_CodeSniffer\Standards\Squiz\Sniffs\WhiteSpace\SuperfluousWhitespaceSniff (line 205) replaced token 9 (T_WHITESPACE) "\t\n$bar" => "\n$bar"
        => Fixing file: 1/1 violations remaining [made 1 pass]...       * fixed 1 violations, starting loop 2 *
        => Fixing file: 0/1 violations remaining [made 2 passes]... DONE in 5ms 
	=> File was overwritten

[jrfnl]

Ah, now we're getting somewhere. I can reproduce the effect when the whitespace is spaces instead of a tab.

Looks like this is a bug in the upstream Generic.WhiteSpace.DisallowSpaceIndent sniff.
I already have a PR open for that sniff upstream. I'll have a look to see if this can be fixed in the same PR or warrants a second one.

[jrfnl]

I've opened a separate PR upstream specifically to address this issue: squizlabs/PHP_CodeSniffer#1702

[jrfnl]

A fix for this has been merged upstream and was included in the PHPCS 3.1.1 release.