Efcore.pg: The CoreRT compiler fails with a StackOverflowException

[YohDeadfall]

From dotnet/corert#6372 by @frankzye.

---

I came an error:

Unhandled Exception: System.InvalidOperationException: A suitable constructor for type 'Npgsql.EntityFrameworkCore.PostgreSQL.Storage.Internal.NpgsqlExecutionStrategyFactory' could not be located. Ensure the type is concrete and services are registered for all parameters of a public constructor.

So I added the following lines to RD.xml:

<Assembly Name="Npgsql.EntityFrameworkCore.PostgreSQL">
  <Type Name="Npgsql.EntityFrameworkCore.PostgreSQL.Storage.Internal.NpgsqlExecutionStrategyFactory" Dynamic="Required All"/>
</Assembly>

But when I compile, it throw an error "Process is terminating due to StackOverflowException", microsoft.dotnet.ilcompiler1.0.0-alpha-26919-01\build\Microsoft.NETCore.Native.targets(200,5): error MSB3073 xxx runtime.win-x64.microsoft.dotnet.ilcompiler1.0.0-alpha-26919-01\tools\ilc" @"objrelease\netcoreapp2.1\win-x64\native\DataExplorer.ilc.rsp exit.

[frankzye]

any update?

[YohDeadfall]

Not yet. Will try to take a look at the upcoming weekend.

[YohDeadfall]

@frankzye Could you post a minimal repro for the issue? Currently I'm getting another error:

error MSB4184: The expression ""System.String[]".GetValue(1)" cannot be evaluated. Index was outside the bounds of the array.

@MichalStrehovsky Is there any way to get the method name which is failed to compile?

[frankzye]

Sample.zip
@YohDeadfall please find the repo, and help to resolve, thanks.

[frankzye]

@YohDeadfall How are things going?

[YohDeadfall]

I had no time to contribute for the last two weeks, sorry.

@roji @austindrenski Would you like to take a look at it?

[roji]

I definitely won't have any bandwidth for this in the next few weeks... Possible afterwards

[austindrenski]

I'm cleaning up the provided sample now to repost as a repo. I haven't worked with the AOT toolchain before, but I'll see what I can do.

_See: https://github.com/austindrenski/EFCore.PG-653_

Related

dotnet/corert#6477

[austindrenski]

So it seems like whatever is triggering this is happening within Database.Migrate(). But commenting out that line reveals a whole sequence of additional EF Core metadata issues.

Thankfully, it looks like some serious legwork has been done in dotnet/corert#6477 by @rubin55 and @MichalStrehovsky to document these issues. I'm working my way though those now before debugging the issue in the migrations code.

[roji]

I'm guessing that many/most of these issues are going to be EF Core-related, rather than specific to Npgsql - it's better to raise these with the EF Core team. I have no idea what the status of EF Core CortRT support is, or whether full CoreRT is a goal of theirs.

[austindrenski]

I'm guessing that many/most of these issues are going to be EF Core-related, rather than specific to Npgsql - it's better to raise these with the EF Core team.

That was my impression after putting in a few hours with it. I would be interested to see a working version, so if anyone makes headway, please keep us in the loop by updating here.

[MichalStrehovsky]

@rubin55 got EF core working in dotnet/corert#6477. It's about finding all the places where EF uses reflection.

[roji]

FWIW if EF Core is made to work, the Npgsql provider shouldn't be much effort beyond that - I don't think we do much that's reflection-based.

[eyupkayadarcin]

Any update ?

[roji]

Not really... https://github.com/dotnet/corert/issues/6477 still seems open even though @MichalStrehovsky mentioned he got EF Core working. The best way forward here is to see whether EF Core works on CoreRT with one of the built-in providers (SqlServer, Sqlient). If they do where Npgsql fails, then we should look into it, otherwise the issue belongs on the EF Core/CoreRT side.

[roji]

Am closing this as a general EF Core issue (https://github.com/dotnet/corert/issues/6477). If, after EF Core works on CoreRT, we find a specific issue with the Npgsql provider (unlikely), we can revisit.