Repository navigation
Need to pull pinning via &MemoryMarshal.GetReference #27308
Description
C# 7.3 allows pinning spans directly. Pinning empty behaves differently than pinning the ref returned from MemoryMarshal.GetReference.
static unsafe void PinMe()
{
int[] emptyArray = new int[0];
Span<int> emptySpan = new Span<int>(emptyArray);
fixed (int* ea = emptyArray)
fixed (int* es = emptySpan)
fixed (int* esr = &MemoryMarshal.GetReference(emptySpan))
{
Console.WriteLine($"Array: 0x{(ulong)ea:X}, Span: 0x{(ulong)es:X}, Address of ref span: 0x{(ulong)esr:X}");
}
}Yields something like:
Array: 0x0, Span: 0x0, Address of ref span: 0x27D97CD1588
cc: @danmosemsft, @jkotas
Activity
danmoseley commentedon Sep 4, 2018
@JeremyKuhne I tried doing this and get a bunch of test failures. Was this what you intended:
https://github.com/dotnet/corefx/compare/master...danmosemsft:memorymarshal?expand=1
Failures like
Any idea what I did wrong?
jkotas commentedon Sep 4, 2018
You have likely discovered real product bugs where the code does not handle empty spans with null pointers correctly. E.g. in the deflater case, I expect that you would see the same problem if you passed in
Span.Emptywith null pointer into the API.JeremyKuhne commentedon Sep 4, 2018
Likely nothing. This is the behavior difference-
&MemoryMarshal.GetReferencewould return a valid pointer.&on an array of zero length returns null. There are likely a few places where we should bail out when we have nothing to do.AlexRadch commentedon Oct 9, 2018
@JeremyKuhne
MemoryMarshal.GetReferencedoes not normalize the pointer for empty Span to null. I don't know if this is a bug or a behavior.You can use
emptySpanthem self or you can use hidden methodemptySpan.GetPinnableReference()joperezr commentedon Jun 5, 2019
@JeremyKuhne is this something we want to do for 3.0? If not I would like to move it to future.
stephentoub commentedon Jun 5, 2019
I'm not entirely sure what this issue is tracking.
JeremyKuhne commentedon Jun 5, 2019
I did not realize it was slightly faster. If it was equivalent there is no reason to use it as it isn't as safe.
Maybe a better issue would be to ensure we have tests that cover "unused" inputs when using GetReference?
joperezr commentedon Jun 5, 2019
From the original issue, I thought the original idea was to do this work for a better perf, but seems like when @danmosemsft attempted it, he flushed out some potential product bugs, so I guess I'm also confused @stephentoub . @JeremyKuhne what is the work that this is really tracking?
JeremyKuhne commentedon Jun 5, 2019
Not using an API that has an unexpected behavior as it is was causing bugs. Pinning empty arrays always returned null, as does directly pinning an empty span.
GetReferencedoes not. Some native APIs require that null is passed for empty arrays (see the linked bug in the description).It was never about performance, I just assumed it wasn't going to negatively impact perf to just pin the span.
joperezr commentedon Jun 21, 2019
I see. @JeremyKuhne do you plan on working on this? If not and this is not critical, we can move this to Future.
2 remaining items
dotnet-policy-service commentedon Jun 14, 2025
Due to lack of recent activity, this issue has been marked as a candidate for backlog cleanup. It will be closed if no further activity occurs within 14 more days. Any new comment (by anyone, not necessarily the author) will undo this process.
This process is part of our issue cleanup automation.
dotnet-policy-service commentedon Jun 28, 2025
This issue will now be closed since it had been marked
no-recent-activitybut received no further activity in the past 14 days. It is still possible to reopen or comment on the issue, but please note that the issue will be locked if it remains inactive for another 30 days.