Add failing test fixture for RemoveDeadReturnRector

[kkmuffme]

Failing Test for RemoveDeadReturnRector

Based on https://getrector.org/demo/38caa690-bf47-405c-af7c-bad1e6d166aa

Exactly the same issue also for \Rector\DeadCode\Rector\ClassMethod\RemoveLastReturnRector::class - do you need a separate test PR for it? (bc it's really exactly the same test code)

[samsonasik]

these kind of mix html+php in function should just be skipped, but if you have better solution, feel free to provide a patch

[kkmuffme]

What's the reason those should be skipped? Bc it's complex to auto-fix those?

EDIT: even if they should be skipped, the rule should skip it by itself.

[samsonasik]

Rector is using nikic/php-parser, which replacing html+php may overap the index.

Still, feel free to provide patch if you found a solution, otherwise, you can use $rectorConfig->skip() option to skip the rule.

[TomasVotruba]

I'll see if Rector know something about the context here, but won't give much home.

In general PHP + HTML are problematic because of switching context.

[TomasVotruba]

So it seems the printer removes <?php and creates invalid code 🤔

Now we have to find out, if that's Rector printer on native php-parser printer :)

[TomasVotruba]

So it seems it's the way php-parser prints the stmts: nikic/PHP-Parser#884

It must be fixed there first to propagate fix here.