[straight-shoota]

This is a trivial refactor, moving the call to StaticArray#to_slice out of the loop. On repeat iterations buf is a Slice so #to_slice is actually a no-op.

However, this change breaks some specs in the interpreter where it produces only gibberish. Looks like stack corruption.
I don't think this change is responsible, it just makes the problem visible.

Related to #14613
Depends on #14615

[BlobCodes]

I don't think this change is responsible, it just makes the problem visible.

The relevant change is probably that before this commit, the whole initial_buf buffer was copied to buf:

crystal/src/crystal/system/unix.cr

Line 5 in f0bd16a

buf = initial_buf

After this, initial_buf was never used again - which means there's 1024 bytes of padding in front of the actually relevant buffer.

It seems like there always has been a stack corruption, but only inside the irrelevant padding region.
Now, the stack corruption inside initial_buf actually matters.