Ensure Array.CreateInstance over reference types works by MichalStrehovsky · Pull Request #100626 · dotnet/runtime

MichalStrehovsky · 2024-04-04T08:20:00Z

We have warning suppressions in place assuming Array.CreateInstance will succeed if it's over reference types, same as MakeGeneric. There's however no analysis to ensure we have the type loader template for __Canon[]. This adds the analysis.

The reflection\GenericAttribute test is currently failing because of this but we don't see it because it's marked RequiresProcessIsolation and we have a bug that we don't look at those tests at all.

We had this bug for a while, but the test regression is more recent because we're getting better at not generating useless stuff. Template for __Canon[] is probably part of any bigger test because it's common.

cc @dotnet/ilc-contrib

We have warning suppressions in place assuming `Array.CreateInstance` will succeed if it's over reference types, same as `MakeGeneric`. There's however no analysis to ensure we have the type loader template for `__Canon[]`. This adds the analysis. The `reflection\GenericAttribute` test is currently failing because of this but we don't see it because it's marked RequiresProcessIsolation and we have a bug that we don't look at those tests at all. We had this bug for a while, but the test regression is more recent because we're getting better at not generating useless stuff. Template for `__Canon[]` is probably part of any bigger test because it's common.

dotnet-policy-service · 2024-04-04T08:20:23Z

Tagging subscribers to this area: @agocke, @MichalStrehovsky, @jkotas
See info in area-owners.md if you want to be subscribed.

Follow up to dotnet#100626. The original fix only worked for unoptimized builds because the list of templates doesn't survive from scanning phase (when we do dataflow) to codegen phase (when we no longer do dataflow). The test that this was supposed to fix is still failing in dotnet#100331.

Follow up to #100626. The original fix only worked for unoptimized builds because the list of templates doesn't survive from scanning phase (when we do dataflow) to codegen phase (when we no longer do dataflow). The test that this was supposed to fix is still failing in #100331.

We have warning suppressions in place assuming `Array.CreateInstance` will succeed if it's over reference types, same as `MakeGeneric`. There's however no analysis to ensure we have the type loader template for `__Canon[]`. This adds the analysis. The `reflection\GenericAttribute` test is currently failing because of this but we don't see it because it's marked RequiresProcessIsolation and we have a bug that we don't look at those tests at all. We had this bug for a while, but the test regression is more recent because we're getting better at not generating useless stuff. Template for `__Canon[]` is probably part of any bigger test because it's common.

Follow up to dotnet#100626. The original fix only worked for unoptimized builds because the list of templates doesn't survive from scanning phase (when we do dataflow) to codegen phase (when we no longer do dataflow). The test that this was supposed to fix is still failing in dotnet#100331.

MichalStrehovsky · 2024-04-04T08:20:00Z

We have warning suppressions in place assuming Array.CreateInstance will succeed if it's over reference types, same as MakeGeneric. There's however no analysis to ensure we have the type loader template for __Canon[]. This adds the analysis.

The reflection\GenericAttribute test is currently failing because of this but we don't see it because it's marked RequiresProcessIsolation and we have a bug that we don't look at those tests at all.

We had this bug for a while, but the test regression is more recent because we're getting better at not generating useless stuff. Template for __Canon[] is probably part of any bigger test because it's common.

cc @dotnet/ilc-contrib

We have warning suppressions in place assuming `Array.CreateInstance` will succeed if it's over reference types, same as `MakeGeneric`. There's however no analysis to ensure we have the type loader template for `__Canon[]`. This adds the analysis. The `reflection\GenericAttribute` test is currently failing because of this but we don't see it because it's marked RequiresProcessIsolation and we have a bug that we don't look at those tests at all. We had this bug for a while, but the test regression is more recent because we're getting better at not generating useless stuff. Template for `__Canon[]` is probably part of any bigger test because it's common.

dotnet-policy-service · 2024-04-04T08:20:23Z

Tagging subscribers to this area: @agocke, @MichalStrehovsky, @jkotas
See info in area-owners.md if you want to be subscribed.

Follow up to dotnet#100626. The original fix only worked for unoptimized builds because the list of templates doesn't survive from scanning phase (when we do dataflow) to codegen phase (when we no longer do dataflow). The test that this was supposed to fix is still failing in dotnet#100331.

Follow up to #100626. The original fix only worked for unoptimized builds because the list of templates doesn't survive from scanning phase (when we do dataflow) to codegen phase (when we no longer do dataflow). The test that this was supposed to fix is still failing in #100331.

We have warning suppressions in place assuming `Array.CreateInstance` will succeed if it's over reference types, same as `MakeGeneric`. There's however no analysis to ensure we have the type loader template for `__Canon[]`. This adds the analysis. The `reflection\GenericAttribute` test is currently failing because of this but we don't see it because it's marked RequiresProcessIsolation and we have a bug that we don't look at those tests at all. We had this bug for a while, but the test regression is more recent because we're getting better at not generating useless stuff. Template for `__Canon[]` is probably part of any bigger test because it's common.

Follow up to dotnet#100626. The original fix only worked for unoptimized builds because the list of templates doesn't survive from scanning phase (when we do dataflow) to codegen phase (when we no longer do dataflow). The test that this was supposed to fix is still failing in dotnet#100331.

MichalStrehovsky · 2024-04-04T08:20:00Z

We have warning suppressions in place assuming Array.CreateInstance will succeed if it's over reference types, same as MakeGeneric. There's however no analysis to ensure we have the type loader template for __Canon[]. This adds the analysis.

The reflection\GenericAttribute test is currently failing because of this but we don't see it because it's marked RequiresProcessIsolation and we have a bug that we don't look at those tests at all.

We had this bug for a while, but the test regression is more recent because we're getting better at not generating useless stuff. Template for __Canon[] is probably part of any bigger test because it's common.

cc @dotnet/ilc-contrib

We have warning suppressions in place assuming `Array.CreateInstance` will succeed if it's over reference types, same as `MakeGeneric`. There's however no analysis to ensure we have the type loader template for `__Canon[]`. This adds the analysis. The `reflection\GenericAttribute` test is currently failing because of this but we don't see it because it's marked RequiresProcessIsolation and we have a bug that we don't look at those tests at all. We had this bug for a while, but the test regression is more recent because we're getting better at not generating useless stuff. Template for `__Canon[]` is probably part of any bigger test because it's common.

dotnet-policy-service · 2024-04-04T08:20:23Z

Tagging subscribers to this area: @agocke, @MichalStrehovsky, @jkotas
See info in area-owners.md if you want to be subscribed.

Follow up to dotnet#100626. The original fix only worked for unoptimized builds because the list of templates doesn't survive from scanning phase (when we do dataflow) to codegen phase (when we no longer do dataflow). The test that this was supposed to fix is still failing in dotnet#100331.

Follow up to #100626. The original fix only worked for unoptimized builds because the list of templates doesn't survive from scanning phase (when we do dataflow) to codegen phase (when we no longer do dataflow). The test that this was supposed to fix is still failing in #100331.

We have warning suppressions in place assuming `Array.CreateInstance` will succeed if it's over reference types, same as `MakeGeneric`. There's however no analysis to ensure we have the type loader template for `__Canon[]`. This adds the analysis. The `reflection\GenericAttribute` test is currently failing because of this but we don't see it because it's marked RequiresProcessIsolation and we have a bug that we don't look at those tests at all. We had this bug for a while, but the test regression is more recent because we're getting better at not generating useless stuff. Template for `__Canon[]` is probably part of any bigger test because it's common.

Follow up to dotnet#100626. The original fix only worked for unoptimized builds because the list of templates doesn't survive from scanning phase (when we do dataflow) to codegen phase (when we no longer do dataflow). The test that this was supposed to fix is still failing in dotnet#100331.

MichalStrehovsky · 2024-04-04T08:20:00Z

We have warning suppressions in place assuming Array.CreateInstance will succeed if it's over reference types, same as MakeGeneric. There's however no analysis to ensure we have the type loader template for __Canon[]. This adds the analysis.

The reflection\GenericAttribute test is currently failing because of this but we don't see it because it's marked RequiresProcessIsolation and we have a bug that we don't look at those tests at all.

We had this bug for a while, but the test regression is more recent because we're getting better at not generating useless stuff. Template for __Canon[] is probably part of any bigger test because it's common.

cc @dotnet/ilc-contrib

We have warning suppressions in place assuming `Array.CreateInstance` will succeed if it's over reference types, same as `MakeGeneric`. There's however no analysis to ensure we have the type loader template for `__Canon[]`. This adds the analysis. The `reflection\GenericAttribute` test is currently failing because of this but we don't see it because it's marked RequiresProcessIsolation and we have a bug that we don't look at those tests at all. We had this bug for a while, but the test regression is more recent because we're getting better at not generating useless stuff. Template for `__Canon[]` is probably part of any bigger test because it's common.

dotnet-policy-service · 2024-04-04T08:20:23Z

Tagging subscribers to this area: @agocke, @MichalStrehovsky, @jkotas
See info in area-owners.md if you want to be subscribed.

Follow up to dotnet#100626. The original fix only worked for unoptimized builds because the list of templates doesn't survive from scanning phase (when we do dataflow) to codegen phase (when we no longer do dataflow). The test that this was supposed to fix is still failing in dotnet#100331.

Follow up to #100626. The original fix only worked for unoptimized builds because the list of templates doesn't survive from scanning phase (when we do dataflow) to codegen phase (when we no longer do dataflow). The test that this was supposed to fix is still failing in #100331.

We have warning suppressions in place assuming `Array.CreateInstance` will succeed if it's over reference types, same as `MakeGeneric`. There's however no analysis to ensure we have the type loader template for `__Canon[]`. This adds the analysis. The `reflection\GenericAttribute` test is currently failing because of this but we don't see it because it's marked RequiresProcessIsolation and we have a bug that we don't look at those tests at all. We had this bug for a while, but the test regression is more recent because we're getting better at not generating useless stuff. Template for `__Canon[]` is probably part of any bigger test because it's common.

Follow up to dotnet#100626. The original fix only worked for unoptimized builds because the list of templates doesn't survive from scanning phase (when we do dataflow) to codegen phase (when we no longer do dataflow). The test that this was supposed to fix is still failing in dotnet#100331.

MichalStrehovsky · 2024-04-04T08:20:00Z

We have warning suppressions in place assuming Array.CreateInstance will succeed if it's over reference types, same as MakeGeneric. There's however no analysis to ensure we have the type loader template for __Canon[]. This adds the analysis.

The reflection\GenericAttribute test is currently failing because of this but we don't see it because it's marked RequiresProcessIsolation and we have a bug that we don't look at those tests at all.

We had this bug for a while, but the test regression is more recent because we're getting better at not generating useless stuff. Template for __Canon[] is probably part of any bigger test because it's common.

cc @dotnet/ilc-contrib

We have warning suppressions in place assuming `Array.CreateInstance` will succeed if it's over reference types, same as `MakeGeneric`. There's however no analysis to ensure we have the type loader template for `__Canon[]`. This adds the analysis. The `reflection\GenericAttribute` test is currently failing because of this but we don't see it because it's marked RequiresProcessIsolation and we have a bug that we don't look at those tests at all. We had this bug for a while, but the test regression is more recent because we're getting better at not generating useless stuff. Template for `__Canon[]` is probably part of any bigger test because it's common.

dotnet-policy-service · 2024-04-04T08:20:23Z

Tagging subscribers to this area: @agocke, @MichalStrehovsky, @jkotas
See info in area-owners.md if you want to be subscribed.

Follow up to dotnet#100626. The original fix only worked for unoptimized builds because the list of templates doesn't survive from scanning phase (when we do dataflow) to codegen phase (when we no longer do dataflow). The test that this was supposed to fix is still failing in dotnet#100331.

Follow up to #100626. The original fix only worked for unoptimized builds because the list of templates doesn't survive from scanning phase (when we do dataflow) to codegen phase (when we no longer do dataflow). The test that this was supposed to fix is still failing in #100331.

We have warning suppressions in place assuming `Array.CreateInstance` will succeed if it's over reference types, same as `MakeGeneric`. There's however no analysis to ensure we have the type loader template for `__Canon[]`. This adds the analysis. The `reflection\GenericAttribute` test is currently failing because of this but we don't see it because it's marked RequiresProcessIsolation and we have a bug that we don't look at those tests at all. We had this bug for a while, but the test regression is more recent because we're getting better at not generating useless stuff. Template for `__Canon[]` is probably part of any bigger test because it's common.

Follow up to dotnet#100626. The original fix only worked for unoptimized builds because the list of templates doesn't survive from scanning phase (when we do dataflow) to codegen phase (when we no longer do dataflow). The test that this was supposed to fix is still failing in dotnet#100331.

MichalStrehovsky · 2024-04-04T08:20:00Z

We have warning suppressions in place assuming Array.CreateInstance will succeed if it's over reference types, same as MakeGeneric. There's however no analysis to ensure we have the type loader template for __Canon[]. This adds the analysis.

The reflection\GenericAttribute test is currently failing because of this but we don't see it because it's marked RequiresProcessIsolation and we have a bug that we don't look at those tests at all.

We had this bug for a while, but the test regression is more recent because we're getting better at not generating useless stuff. Template for __Canon[] is probably part of any bigger test because it's common.

cc @dotnet/ilc-contrib

We have warning suppressions in place assuming `Array.CreateInstance` will succeed if it's over reference types, same as `MakeGeneric`. There's however no analysis to ensure we have the type loader template for `__Canon[]`. This adds the analysis. The `reflection\GenericAttribute` test is currently failing because of this but we don't see it because it's marked RequiresProcessIsolation and we have a bug that we don't look at those tests at all. We had this bug for a while, but the test regression is more recent because we're getting better at not generating useless stuff. Template for `__Canon[]` is probably part of any bigger test because it's common.

dotnet-policy-service · 2024-04-04T08:20:23Z

Tagging subscribers to this area: @agocke, @MichalStrehovsky, @jkotas
See info in area-owners.md if you want to be subscribed.

Follow up to dotnet#100626. The original fix only worked for unoptimized builds because the list of templates doesn't survive from scanning phase (when we do dataflow) to codegen phase (when we no longer do dataflow). The test that this was supposed to fix is still failing in dotnet#100331.

Follow up to #100626. The original fix only worked for unoptimized builds because the list of templates doesn't survive from scanning phase (when we do dataflow) to codegen phase (when we no longer do dataflow). The test that this was supposed to fix is still failing in #100331.

We have warning suppressions in place assuming `Array.CreateInstance` will succeed if it's over reference types, same as `MakeGeneric`. There's however no analysis to ensure we have the type loader template for `__Canon[]`. This adds the analysis. The `reflection\GenericAttribute` test is currently failing because of this but we don't see it because it's marked RequiresProcessIsolation and we have a bug that we don't look at those tests at all. We had this bug for a while, but the test regression is more recent because we're getting better at not generating useless stuff. Template for `__Canon[]` is probably part of any bigger test because it's common.

Follow up to dotnet#100626. The original fix only worked for unoptimized builds because the list of templates doesn't survive from scanning phase (when we do dataflow) to codegen phase (when we no longer do dataflow). The test that this was supposed to fix is still failing in dotnet#100331.

MichalStrehovsky · 2024-04-04T08:20:00Z

We have warning suppressions in place assuming Array.CreateInstance will succeed if it's over reference types, same as MakeGeneric. There's however no analysis to ensure we have the type loader template for __Canon[]. This adds the analysis.

The reflection\GenericAttribute test is currently failing because of this but we don't see it because it's marked RequiresProcessIsolation and we have a bug that we don't look at those tests at all.

We had this bug for a while, but the test regression is more recent because we're getting better at not generating useless stuff. Template for __Canon[] is probably part of any bigger test because it's common.

cc @dotnet/ilc-contrib

We have warning suppressions in place assuming `Array.CreateInstance` will succeed if it's over reference types, same as `MakeGeneric`. There's however no analysis to ensure we have the type loader template for `__Canon[]`. This adds the analysis. The `reflection\GenericAttribute` test is currently failing because of this but we don't see it because it's marked RequiresProcessIsolation and we have a bug that we don't look at those tests at all. We had this bug for a while, but the test regression is more recent because we're getting better at not generating useless stuff. Template for `__Canon[]` is probably part of any bigger test because it's common.

dotnet-policy-service · 2024-04-04T08:20:23Z

Tagging subscribers to this area: @agocke, @MichalStrehovsky, @jkotas
See info in area-owners.md if you want to be subscribed.

Follow up to dotnet#100626. The original fix only worked for unoptimized builds because the list of templates doesn't survive from scanning phase (when we do dataflow) to codegen phase (when we no longer do dataflow). The test that this was supposed to fix is still failing in dotnet#100331.

Follow up to #100626. The original fix only worked for unoptimized builds because the list of templates doesn't survive from scanning phase (when we do dataflow) to codegen phase (when we no longer do dataflow). The test that this was supposed to fix is still failing in #100331.

We have warning suppressions in place assuming `Array.CreateInstance` will succeed if it's over reference types, same as `MakeGeneric`. There's however no analysis to ensure we have the type loader template for `__Canon[]`. This adds the analysis. The `reflection\GenericAttribute` test is currently failing because of this but we don't see it because it's marked RequiresProcessIsolation and we have a bug that we don't look at those tests at all. We had this bug for a while, but the test regression is more recent because we're getting better at not generating useless stuff. Template for `__Canon[]` is probably part of any bigger test because it's common.

Follow up to dotnet#100626. The original fix only worked for unoptimized builds because the list of templates doesn't survive from scanning phase (when we do dataflow) to codegen phase (when we no longer do dataflow). The test that this was supposed to fix is still failing in dotnet#100331.

MichalStrehovsky · 2024-04-04T08:20:00Z

We have warning suppressions in place assuming Array.CreateInstance will succeed if it's over reference types, same as MakeGeneric. There's however no analysis to ensure we have the type loader template for __Canon[]. This adds the analysis.

The reflection\GenericAttribute test is currently failing because of this but we don't see it because it's marked RequiresProcessIsolation and we have a bug that we don't look at those tests at all.

We had this bug for a while, but the test regression is more recent because we're getting better at not generating useless stuff. Template for __Canon[] is probably part of any bigger test because it's common.

cc @dotnet/ilc-contrib

We have warning suppressions in place assuming `Array.CreateInstance` will succeed if it's over reference types, same as `MakeGeneric`. There's however no analysis to ensure we have the type loader template for `__Canon[]`. This adds the analysis. The `reflection\GenericAttribute` test is currently failing because of this but we don't see it because it's marked RequiresProcessIsolation and we have a bug that we don't look at those tests at all. We had this bug for a while, but the test regression is more recent because we're getting better at not generating useless stuff. Template for `__Canon[]` is probably part of any bigger test because it's common.

dotnet-policy-service · 2024-04-04T08:20:23Z

Tagging subscribers to this area: @agocke, @MichalStrehovsky, @jkotas
See info in area-owners.md if you want to be subscribed.

Follow up to dotnet#100626. The original fix only worked for unoptimized builds because the list of templates doesn't survive from scanning phase (when we do dataflow) to codegen phase (when we no longer do dataflow). The test that this was supposed to fix is still failing in dotnet#100331.

Follow up to #100626. The original fix only worked for unoptimized builds because the list of templates doesn't survive from scanning phase (when we do dataflow) to codegen phase (when we no longer do dataflow). The test that this was supposed to fix is still failing in #100331.

We have warning suppressions in place assuming `Array.CreateInstance` will succeed if it's over reference types, same as `MakeGeneric`. There's however no analysis to ensure we have the type loader template for `__Canon[]`. This adds the analysis. The `reflection\GenericAttribute` test is currently failing because of this but we don't see it because it's marked RequiresProcessIsolation and we have a bug that we don't look at those tests at all. We had this bug for a while, but the test regression is more recent because we're getting better at not generating useless stuff. Template for `__Canon[]` is probably part of any bigger test because it's common.

Follow up to dotnet#100626. The original fix only worked for unoptimized builds because the list of templates doesn't survive from scanning phase (when we do dataflow) to codegen phase (when we no longer do dataflow). The test that this was supposed to fix is still failing in dotnet#100331.

MichalStrehovsky requested a review from sbomer April 4, 2024 08:20

MichalStrehovsky requested a review from marek-safar as a code owner April 4, 2024 08:20

ghost added the area-NativeAOT-coreclr label Apr 4, 2024

dotnet-policy-serviceBot assigned MichalStrehovsky Apr 4, 2024

jkotas approved these changes Apr 4, 2024
View reviewed changes

MichalStrehovsky merged commit beac274 into dotnet:mainApr 5, 2024

MichalStrehovsky deleted the arraycreate branch April 5, 2024 05:19

MichalStrehovsky mentioned this pull request Apr 15, 2024
Ensure Array.CreateInstance for reference types works #101048
Merged

MichalStrehovsky mentioned this pull request Apr 17, 2024
[net9.0-preview3]: NativeAOT System.Text.Json throws "System.NotSupportedException: 'System.Text.Json.Serialization.JsonDerivedTypeAttribute[]' is missing native code or metadata." #101205
Closed

github-actionsBot locked and limited conversation to collaborators May 5, 2024