I am not sure that this qualifies as a bug fix. Reading the code it clear that the constructor expects a list of extensions not a single one. Maybe we should clarify that by adding a docblock?

Like this:

/**
 * @param array<string, array<int, string>> $map
 */
public function __construct(array $map = [])
{
    // ...
}

@xabbuh
The bugfix is on line 56, right now it's setting the mimetype for an extension as a string and not an array. getMimeTypes is expected to return an array but when it gets the extension it's a string and throws an error.

What I wanted to say is that this was probably not supposed to work:

$mt = new MimeTypes([
    'text/bar' => 'foo',
]);

It should have been written like this instead:

$mt = new MimeTypes([
    'text/bar' => ['foo'],
]);

What's wrong with adding that in?

There's nothing wrong with it. I am just questioning whether this has been an intended usage. Depending on the answer this patch would either be a bugfix or a new feature.

@xabbuh I have removed the line of convenience code you're so caught up on. Now it's only a bug fix.

@nicolas-grekas I think the 7.1 integration just needs to be rerun, some sort of hiccup, probably because of 2 commits back to back.

At least there is a bug that if you pass a string, it explodes.
I'm fine merging as a bugfix for this reason on my side.

@xabbuh I have removed the line of convenience code you're so caught up on. Now it's only a bug fix.

I think there is a misunderstanding here. My concern was not if changing that line is important, but whether or not we want to treat the whole patch as a bugfix or as a new feature. I can see good reasons for both that's why I wanted to discuss that.

I misunderstood the comment I replied to. I understand the issue now and the proposed change fixes it. 👍

Thank you @mcneely.