Weird PublishAot memory differences without actually publishing

[ThomasFOG]

Description

Hello,

This is really puzzling in a way that it makes me doubt of myself.

While debugging from Visual Studio in DEBUG configuration, apps don't have the same memory behavior depending on whether PublishAot is set (which shouldn't even trigger outside dotnet publish, right?) and on whether the debugger is attached.

While debugging and PublishAot isn't set, memory will keep building up like if there were memory allocations going on.

While debugging and PublisAot is set, memory doesn't build up and remains stable if we don't allocate anything.

When the same exe is run without the debugger, memory is stable no matter what. This only happens when the debugger is attached and PublishAot isn't set.

Don't know if this has something to do with some runtime analyzers?

Happens in both VS 17.13.2 and 17.12.

VS components installed:

• Visual C++ 2022 00482-90000-00000-AA134
• Microsoft Visual C++ 2022
• ASP.NET and Web Tools 17.13.124.35287
• ASP.NET and Web Tools
• Azure App Service Tools v3.0.0 17.13.124.35287
• Azure App Service Tools v3.0.0
• C# Tools 4.13.0-3.25112.7+afe5bc4a827030229ab79ec596b3cb91d4e87872
• Microsoft JVM Debugger 1.0
• Mono Debugging for Visual Studio 17.13.14 (1c94ab9)
• NuGet Package Manager 6.13.2
• Razor (ASP.NET Core) 17.13.3.2511303+2405f884fd9c848997c6be0607b18b55544f526a
• TypeScript Tools 17.0.31211.2001
• Visual Basic Tools 4.13.0-3.25112.7+afe5bc4a827030229ab79ec596b3cb91d4e87872
• Visual F# Tools 17.13.0-beta.25056.5+63a09289745da5c256e7baf5f4194a750b1ec878
• Visual Studio IntelliCode 2.2
• VisualStudio.DeviceLog 1.0
• VisualStudio.Mac 1.0
• VSPackage Extension 1.0
• Xamarin 17.13.0.171 (d17.13@5b137a5)
• Xamarin Designer 17.13.3.4 (remotes/origin/main@59e9def289)
• Xamarin.Android SDK 13.2.2.0 (d17-5/45b0e14)

Reproduction Steps

Debug from Visual Studio the most simple app and add/remove PublishAot.

<Project Sdk="Microsoft.NET.Sdk">

  <PropertyGroup>
    <OutputType>Exe</OutputType>
    <TargetFramework>net9.0</TargetFramework>
    <ImplicitUsings>enable</ImplicitUsings>
  </PropertyGroup>

</Project>

Console.WriteLine("Hello, World!");

long test = System.GC.GetTotalMemory(false);

while (true)
{
    long test2 = System.GC.GetTotalMemory(false);

    if (test2 != test)
        System.Diagnostics.Debug.WriteLine(test2 - test);
}

Expected behavior

Memory behavior should stay the same?

Actual behavior

Memory behavior should stay the same?

Regression?

No response

Known Workarounds

No response

Configuration

Tested on .NET 9.0.103, 9.0.200, and 8.0.406.

Other information

No response

[dotnet-policy-service]

Tagging subscribers to this area: @tommcdon
See info in area-owners.md if you want to be subscribed.

[vitek-karas]

PublishAot is set (which shouldn't even trigger outside dotnet publish, right?)

PublishAot=true modifies some feature switches so the runtime behavior is different. The goal is that the runtime behavior should be the same in debug and release (so without AOT and with AOT). It's possible that one of these affects memory behavior, but I don't know if that's the case.