Runtime: Consider documenting the initial contents of the Span<char> passed into SpanAction of string.Create method

[Gnbrkm41]

Looking at the documents It is currently unclear on whether the contents are zeroed or contain random data. I believe currently the contents are zeroed (returned from FastAllocateString) but it would be nice to specify whether it is something people can rely on or not.

[Dotnet-GitSync-Bot]

I couldn't figure out the best area label to add to this issue. Please help me learn by adding exactly one area label.

[bartonjs]

I believe the answer is "it's completely not guaranteed, you're responsible for assigning every position"

[Gnbrkm41]

I did guess so. However could it be a breaking change if we were to change its implementation to give non-zeroed string? It has not been documented so far and currently it zeroes. It wouldn't have been a breaking change if it were documented to suggest it could contain gibberish data (like stackalloc etc) when it was first introduced.

[danmosemsft]

@jkotas would it be reasonable to document this?

[jkotas]

It's completely not guaranteed, you're responsible for assigning every position

Yes, it would be a good idea to mention this in the documentation.

could it be a breaking change if we were to change its implementation to give non-zeroed string

We would not change this in servicing where we try to maintain bug-for-bug compatibility. I believe that we would be comfortable to change it for major version where we try to be compatible, but not bug-for-bug compatible.

[Gnbrkm41]

Great! I can put up a PR on the docs repo in the morning (local time).

Though, do I need to make two separate PRs (one in the docs repo, one in the runtime repo), or do the changes made in XML comments propagate down to the dotnet-api-docs repo?

[jkotas]

The XML comments and the docs repo are not kept in sync automatically. It has to be done manually.

[bartonjs]

I'd just do it in Remarks on the docs site, I wouldn't do anything here.

@jkotas So zero-init memory is guaranteed here for 2.1-5.0?

[jkotas]

It always happens to zero-init in 2.1 - 3.1. We have not shipped 5.0 yet...

[stephentoub]

We should document that it's not guaranteed. I believe that's what Jan is saying as well. Correct, Jan?

[jkotas]

Right. The docs should not mention that it happens to zero-init in 2.1 - 3.1

[bartonjs]

Somehow I decided Jan had quoted the "it's zero" instead of my "it's not guaranteed". Good clarification.