[HttpFoundation] Failed to set session handler, headers already sent by phpunit

[anlutro]

Q	A
Bug report?	yes
Feature request?	no
BC Break report?	yes
RFC?	no
Symfony version	3.4.x

I have a project that depends on symfony/http-foundation. With 3.4.x-dev I'm getting some RuntimeErrors recently:

RuntimeException: Failed to set the session handler because headers have already been sent
by ".../vendor/phpunit/phpunit/src/Util/Printer.php" at line 110.

.../vendor/symfony/http-foundation/Session/Storage/NativeSessionStorage.php:397
.../vendor/symfony/http-foundation/Session/Storage/NativeSessionStorage.php:112
.../classes/Http/SessionProvider.php:128
.../vendor/autarky/container/classes/Factory/Factory.php:89
.../vendor/autarky/container/classes/Container.php:256
.../vendor/autarky/container/classes/Container.php:226
.../classes/Application.php:511
.../tests/unit/Session/SessionProviderTest.php:20

I tracked down the line to #23711 which was merged 2 days ago. If I remove the if headers_sent thing from setSaveHandler, my test runs fine again.

PHPUnit 5.7.22, symfony/http-foundation is at 0fe9b3d

[nicolas-grekas]

You're going to have this error anyway when running on PHP 7.2, so you should fix your tests.
Running them in a separate process should do the trick.

[sroze]

@nicolas-grekas I think the issue is a bit deeper. What's new in PHP 7.2 is that PHP won't allow you anymore to change the session's ini settings or save handlers after the session has been started.

This should be fixed in #24516 (i.e. you will get an exception only if you try to change things while creating the NativeSessionStorage object, not every time you create an instance of it 🤞) - but it's blocked by a remaining failing test.

[garak]

This should be merged in 3.4 branch too. I'm on a project with php 7.2/Symfony 3.4-beta3 and there's no way to get this working 😞

[garak]

Also, I tried a dirty hack, by copying NativeSessionStorage class from #24531, and I'm still getting the same error ("Failed to set...")

[umpirsky]

Upgraded to v3.4.0-BETA3, same error I reported in #24800, which is probably a duplicate of this issue.

@nicolas-grekas I think we should reopen.