Support UTF-8 BOM stripping in TextInputStream

[slqy123]

Currently, TextInputStream treats all text as raw bytes and does not strip UTF-8 BOM (EF BB BF). Cue sheets saved with UTF-8 BOM can fail to parse the first line because the first token is prefixed with BOM bytes.

This PR Strips UTF-8 BOM in TextInputStream constructor. It tries to read the first 3 bytes into buffer, and consume those bytes if a BOM is present.

[MaxKellermann]

I don't think that's the right place to do this. This makes the constructor potentially blocking (and out of the caller's control), and the constructor can now throw on I/O error (which will crash MPD). I find it rather surprising that the constructor will implicitly do blocking I/O. Plus, it may deadlock if the caller happens to already hold a mutex lock.

[slqy123]

I don't think that's the right place to do this. This makes the constructor potentially blocking (and out of the caller's control), and the constructor can now throw on I/O error (which will crash MPD). I find it rather surprising that the constructor will implicitly do blocking I/O. Plus, it may deadlock if the caller happens to already hold a mutex lock.

Thank you for the feedback. I will move this to TextInputStream::ReadLine and add a flag to track if the BOM was checked.

[MaxKellermann]

That's much better. Now please amend the commit message - it contains less than the PR text, and that's unfortunate, because PR text gets lost on GitHub's proprietary website.
If this fixes a problem users have currently, it should probably go to the stable branch v0.24.x, but I can easily cherry-pick it over there.

[slqy123]

If this fixes a problem users have currently, it should probably go to the stable branch v0.24.x, but I can easily cherry-pick it over there.

I didn't find an existing issue for this.
Since it only breaks the first line, most of the cuesheet still parses. it's easy to miss, but users are still affected. So I believe it should go to the stable branch.

[MaxKellermann]

You edited the commit message from
"input/TextInputStream: strip UTF-8 BOM"
to
"input/TextInputStream: support UTF-8 BOM stripping in TextInputStream"

This is still

• less information than the PR
• not more than before; it just adds the (weasel) word "support" and mentions TextInputStream twice, but that doesn't add any information

[slqy123]

You edited the commit message from "input/TextInputStream: strip UTF-8 BOM" to "input/TextInputStream: support UTF-8 BOM stripping in TextInputStream"

This is still

* less information than the PR

* not more than before; it just adds the (weasel) word "support" and mentions TextInputStream twice, but that doesn't add any information

OK, I have updated my commit message with a detailed explaination

[MaxKellermann]

Cherry-picked to v0.24.x: 98bb249