Release Android X509 chain certificate GREFs by simonrozsival · Pull Request #128284 · dotnet/runtime

simonrozsival · 2026-05-16T13:01:15Z

Fixes a JNI global-reference leak in the Android X509 chain PAL.

AndroidCryptoNative_X509ChainGetCertificates returns caller-owned JNI global refs for the certificates in the built chain. Constructing X509Certificate2 from those pointers duplicates the refs for managed ownership, but the original returned refs were never released. This releases those temporary refs after the managed certificates have been created.

Regression test

BuildChainRepeatedly_DoesNotExhaustGlobalReferences is a new [OuterLoop] Android-only test that builds 8,600 6-certificate PKI chains via CertificateAuthority.BuildPrivatePki. Without this PR each successful build leaks 6 JNI global references, so 8,600 iterations would leak 51,600 — past Android's default 51,200 entry limit.

Validation

Run on a local Android emulator (API 36, arm64-v8a, Apple Silicon host, Mono interpreter):

With this PR's fix: ✅ passes (xharness exit 0) in ~9.5 minutes.
Without the fix (verified by reverting the new try/finally block in Interop.X509Chain.cs): ❌ crashes between iter ~7,750 and ~8,500 with JNI ERROR (app bug): global reference table overflow (max=51200) and a tombstone; xharness exit 80 (APP_CRASH).

The Android X509 chain PAL returns JNI global references for chain certificates. Creating X509Certificate2 from those pointers duplicates the references for managed ownership, leaving the native-returned references caller-owned. Release those temporary references after conversion so repeated chain builds do not exhaust ART global refs. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

dotnet-policy-service · 2026-05-16T13:12:05Z

Tagging subscribers to this area: @dotnet/ncl, @bartonjs, @vcsjones
See info in area-owners.md if you want to be subscribed.

Copilot

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.

…ref limit Adds [OuterLoop] regression test BuildChainRepeatedly_DoesNotExhaustGlobalReferences that builds a 6-certificate chain (root + 4 intermediates + endCert) 8,600 times. Without the gref-release fix in PR dotnet#128284, AndroidCryptoNative_X509ChainGetCertificates returns caller-owned JNI global references that X509Certificate2 then duplicates, leaving the native-returned refs orphaned. A 6-cert chain leaks 6 grefs per successful build; 8,600 builds therefore leak 51,600 references, which exceeds Android's default global-reference table limit (51,200) and aborts the process with 'global reference table overflow (max=51200)'. Threshold validation against the unfixed code path: - 8,400 iterations completed successfully - 8,500 iterations crashed with the gref-table overflow With the managed try/finally cleanup in place (Interop.X509Chain.X509ChainGetCertificates), 8,600 iterations complete cleanly. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…ref limit Adds [OuterLoop] companion test BuildChainRepeatedlyOnFailure_DoesNotExhaustGlobalReferences that builds a 6-certificate chain and forces it to fail (VerificationTime far in the future, well outside cert validity) 60,000 times. The success-path test (BuildChainRepeatedly_DoesNotExhaustGlobalReferences) exercises X509ChainGetCertificates, which PR dotnet#128284 fixed. The failure path is a different code path: managed code skips GetCertificates entirely when chain.Build returns false, and instead invokes X509ChainGetErrors and walks the resulting error list. The pal_x509chain.c side allocates xmalloc'd UTF-16 message buffers per error; the managed side is expected to Marshal.FreeHGlobal each one. The failure path doesn't currently leak — the test is defensive against future regressions that introduce a per-iteration leak on this path (e.g., a forgotten Marshal.FreeHGlobal, a missing ReleaseGRef on a throwable, or a regression in X509ChainBuild's exception-handling flow). 60,000 iterations is chosen to overflow the 51,200-entry JNI global reference table even under a worst-case 1-gref-per-iteration regression. The 6-cert chain mirrors the success-path test so the two tests cover the same chain shape. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Copilot

Pull request overview

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

simonrozsival · 2026-05-19T21:13:24Z

/backport to release/10.0

github-actions · 2026-05-19T21:13:34Z

Started backporting to release/10.0 (link to workflow run)

@simonrozsival

Backport of #128284 to release/10.0 /cc @simonrozsival ## Customer Impact - [x] Customer reported - [ ] Found internally A customer reported that their app was crashing due to a memory leak in X509Chain class (JNI global ref table overflow). ## Regression - [ ] Yes - [x] No [If yes, specify when the regression was introduced. Provide the PR or commit if known.] ## Testing The new unit test (ChainTests.BuildChainRepeatedly_DoesNotExhaustGlobalReferences) passes locally on Android emulator. ## Risk Low. The change only affects Android and the code path is well covered by tests. **IMPORTANT**: If this backport is for a servicing release, please verify that: - For .NET 8 and .NET 9: The PR target branch is `release/X.0-staging`, not `release/X.0`. - For .NET 10+: The PR target branch is `release/X.0` (no `-staging` suffix). ## Package authoring no longer needed in .NET 9 **IMPORTANT**: Starting with .NET 9, you no longer need to edit a NuGet package's csproj to enable building and bump the version. Keep in mind that we still need package authoring in .NET 8 and older versions. --------- Co-authored-by: Simon Rozsival <simon@rozsival.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

@jkotas

Follow-up to #128284 Follow-up to #128385 This PR removes an unnecessary long running unit test added in #128284 /cc @jkotas

Fixes Android X509Store PAL cleanup paths for certificate/private-key entries and JNI local references to avoid memory leaks. Follow-up to #128284 ## Changes - Dispose the Android `KeyStore.PrivateKeyEntry` wrapper held by `AndroidCertificatePal`. - Release JNI local references on Android X509Store cleanup paths: - trusted certificate enumeration - default store open failure/success cleanup - remove-certificate early success path - Add JNI exception checks when advancing Android KeyStore alias enumerations. --------- Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

simonrozsival · 2026-05-16T13:01:15Z

Fixes a JNI global-reference leak in the Android X509 chain PAL.

AndroidCryptoNative_X509ChainGetCertificates returns caller-owned JNI global refs for the certificates in the built chain. Constructing X509Certificate2 from those pointers duplicates the refs for managed ownership, but the original returned refs were never released. This releases those temporary refs after the managed certificates have been created.

Regression test

BuildChainRepeatedly_DoesNotExhaustGlobalReferences is a new [OuterLoop] Android-only test that builds 8,600 6-certificate PKI chains via CertificateAuthority.BuildPrivatePki. Without this PR each successful build leaks 6 JNI global references, so 8,600 iterations would leak 51,600 — past Android's default 51,200 entry limit.

Validation

Run on a local Android emulator (API 36, arm64-v8a, Apple Silicon host, Mono interpreter):

With this PR's fix: ✅ passes (xharness exit 0) in ~9.5 minutes.
Without the fix (verified by reverting the new try/finally block in Interop.X509Chain.cs): ❌ crashes between iter ~7,750 and ~8,500 with JNI ERROR (app bug): global reference table overflow (max=51200) and a tombstone; xharness exit 80 (APP_CRASH).

The Android X509 chain PAL returns JNI global references for chain certificates. Creating X509Certificate2 from those pointers duplicates the references for managed ownership, leaving the native-returned references caller-owned. Release those temporary references after conversion so repeated chain builds do not exhaust ART global refs. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

dotnet-policy-service · 2026-05-16T13:12:05Z

Tagging subscribers to this area: @dotnet/ncl, @bartonjs, @vcsjones
See info in area-owners.md if you want to be subscribed.

Copilot

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.

…ref limit Adds [OuterLoop] regression test BuildChainRepeatedly_DoesNotExhaustGlobalReferences that builds a 6-certificate chain (root + 4 intermediates + endCert) 8,600 times. Without the gref-release fix in PR dotnet#128284, AndroidCryptoNative_X509ChainGetCertificates returns caller-owned JNI global references that X509Certificate2 then duplicates, leaving the native-returned refs orphaned. A 6-cert chain leaks 6 grefs per successful build; 8,600 builds therefore leak 51,600 references, which exceeds Android's default global-reference table limit (51,200) and aborts the process with 'global reference table overflow (max=51200)'. Threshold validation against the unfixed code path: - 8,400 iterations completed successfully - 8,500 iterations crashed with the gref-table overflow With the managed try/finally cleanup in place (Interop.X509Chain.X509ChainGetCertificates), 8,600 iterations complete cleanly. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…ref limit Adds [OuterLoop] companion test BuildChainRepeatedlyOnFailure_DoesNotExhaustGlobalReferences that builds a 6-certificate chain and forces it to fail (VerificationTime far in the future, well outside cert validity) 60,000 times. The success-path test (BuildChainRepeatedly_DoesNotExhaustGlobalReferences) exercises X509ChainGetCertificates, which PR dotnet#128284 fixed. The failure path is a different code path: managed code skips GetCertificates entirely when chain.Build returns false, and instead invokes X509ChainGetErrors and walks the resulting error list. The pal_x509chain.c side allocates xmalloc'd UTF-16 message buffers per error; the managed side is expected to Marshal.FreeHGlobal each one. The failure path doesn't currently leak — the test is defensive against future regressions that introduce a per-iteration leak on this path (e.g., a forgotten Marshal.FreeHGlobal, a missing ReleaseGRef on a throwable, or a regression in X509ChainBuild's exception-handling flow). 60,000 iterations is chosen to overflow the 51,200-entry JNI global reference table even under a worst-case 1-gref-per-iteration regression. The 6-cert chain mirrors the success-path test so the two tests cover the same chain shape. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Copilot

Pull request overview

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

simonrozsival · 2026-05-19T21:13:24Z

/backport to release/10.0

github-actions · 2026-05-19T21:13:34Z

Started backporting to release/10.0 (link to workflow run)

@simonrozsival

Backport of #128284 to release/10.0 /cc @simonrozsival ## Customer Impact - [x] Customer reported - [ ] Found internally A customer reported that their app was crashing due to a memory leak in X509Chain class (JNI global ref table overflow). ## Regression - [ ] Yes - [x] No [If yes, specify when the regression was introduced. Provide the PR or commit if known.] ## Testing The new unit test (ChainTests.BuildChainRepeatedly_DoesNotExhaustGlobalReferences) passes locally on Android emulator. ## Risk Low. The change only affects Android and the code path is well covered by tests. **IMPORTANT**: If this backport is for a servicing release, please verify that: - For .NET 8 and .NET 9: The PR target branch is `release/X.0-staging`, not `release/X.0`. - For .NET 10+: The PR target branch is `release/X.0` (no `-staging` suffix). ## Package authoring no longer needed in .NET 9 **IMPORTANT**: Starting with .NET 9, you no longer need to edit a NuGet package's csproj to enable building and bump the version. Keep in mind that we still need package authoring in .NET 8 and older versions. --------- Co-authored-by: Simon Rozsival <simon@rozsival.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

@jkotas

Follow-up to #128284 Follow-up to #128385 This PR removes an unnecessary long running unit test added in #128284 /cc @jkotas

Fixes Android X509Store PAL cleanup paths for certificate/private-key entries and JNI local references to avoid memory leaks. Follow-up to #128284 ## Changes - Dispose the Android `KeyStore.PrivateKeyEntry` wrapper held by `AndroidCertificatePal`. - Release JNI local references on Android X509Store cleanup paths: - trusted certificate enumeration - default store open failure/success cleanup - remove-certificate early success path - Add JNI exception checks when advancing Android KeyStore alias enumerations. --------- Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

simonrozsival · 2026-05-16T13:01:15Z

Fixes a JNI global-reference leak in the Android X509 chain PAL.

AndroidCryptoNative_X509ChainGetCertificates returns caller-owned JNI global refs for the certificates in the built chain. Constructing X509Certificate2 from those pointers duplicates the refs for managed ownership, but the original returned refs were never released. This releases those temporary refs after the managed certificates have been created.

Regression test

BuildChainRepeatedly_DoesNotExhaustGlobalReferences is a new [OuterLoop] Android-only test that builds 8,600 6-certificate PKI chains via CertificateAuthority.BuildPrivatePki. Without this PR each successful build leaks 6 JNI global references, so 8,600 iterations would leak 51,600 — past Android's default 51,200 entry limit.

Validation

Run on a local Android emulator (API 36, arm64-v8a, Apple Silicon host, Mono interpreter):

With this PR's fix: ✅ passes (xharness exit 0) in ~9.5 minutes.
Without the fix (verified by reverting the new try/finally block in Interop.X509Chain.cs): ❌ crashes between iter ~7,750 and ~8,500 with JNI ERROR (app bug): global reference table overflow (max=51200) and a tombstone; xharness exit 80 (APP_CRASH).

The Android X509 chain PAL returns JNI global references for chain certificates. Creating X509Certificate2 from those pointers duplicates the references for managed ownership, leaving the native-returned references caller-owned. Release those temporary references after conversion so repeated chain builds do not exhaust ART global refs. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

dotnet-policy-service · 2026-05-16T13:12:05Z

Tagging subscribers to this area: @dotnet/ncl, @bartonjs, @vcsjones
See info in area-owners.md if you want to be subscribed.

Copilot

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.

…ref limit Adds [OuterLoop] regression test BuildChainRepeatedly_DoesNotExhaustGlobalReferences that builds a 6-certificate chain (root + 4 intermediates + endCert) 8,600 times. Without the gref-release fix in PR dotnet#128284, AndroidCryptoNative_X509ChainGetCertificates returns caller-owned JNI global references that X509Certificate2 then duplicates, leaving the native-returned refs orphaned. A 6-cert chain leaks 6 grefs per successful build; 8,600 builds therefore leak 51,600 references, which exceeds Android's default global-reference table limit (51,200) and aborts the process with 'global reference table overflow (max=51200)'. Threshold validation against the unfixed code path: - 8,400 iterations completed successfully - 8,500 iterations crashed with the gref-table overflow With the managed try/finally cleanup in place (Interop.X509Chain.X509ChainGetCertificates), 8,600 iterations complete cleanly. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…ref limit Adds [OuterLoop] companion test BuildChainRepeatedlyOnFailure_DoesNotExhaustGlobalReferences that builds a 6-certificate chain and forces it to fail (VerificationTime far in the future, well outside cert validity) 60,000 times. The success-path test (BuildChainRepeatedly_DoesNotExhaustGlobalReferences) exercises X509ChainGetCertificates, which PR dotnet#128284 fixed. The failure path is a different code path: managed code skips GetCertificates entirely when chain.Build returns false, and instead invokes X509ChainGetErrors and walks the resulting error list. The pal_x509chain.c side allocates xmalloc'd UTF-16 message buffers per error; the managed side is expected to Marshal.FreeHGlobal each one. The failure path doesn't currently leak — the test is defensive against future regressions that introduce a per-iteration leak on this path (e.g., a forgotten Marshal.FreeHGlobal, a missing ReleaseGRef on a throwable, or a regression in X509ChainBuild's exception-handling flow). 60,000 iterations is chosen to overflow the 51,200-entry JNI global reference table even under a worst-case 1-gref-per-iteration regression. The 6-cert chain mirrors the success-path test so the two tests cover the same chain shape. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Copilot

Pull request overview

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

simonrozsival · 2026-05-19T21:13:24Z

/backport to release/10.0

github-actions · 2026-05-19T21:13:34Z

Started backporting to release/10.0 (link to workflow run)

@simonrozsival

Backport of #128284 to release/10.0 /cc @simonrozsival ## Customer Impact - [x] Customer reported - [ ] Found internally A customer reported that their app was crashing due to a memory leak in X509Chain class (JNI global ref table overflow). ## Regression - [ ] Yes - [x] No [If yes, specify when the regression was introduced. Provide the PR or commit if known.] ## Testing The new unit test (ChainTests.BuildChainRepeatedly_DoesNotExhaustGlobalReferences) passes locally on Android emulator. ## Risk Low. The change only affects Android and the code path is well covered by tests. **IMPORTANT**: If this backport is for a servicing release, please verify that: - For .NET 8 and .NET 9: The PR target branch is `release/X.0-staging`, not `release/X.0`. - For .NET 10+: The PR target branch is `release/X.0` (no `-staging` suffix). ## Package authoring no longer needed in .NET 9 **IMPORTANT**: Starting with .NET 9, you no longer need to edit a NuGet package's csproj to enable building and bump the version. Keep in mind that we still need package authoring in .NET 8 and older versions. --------- Co-authored-by: Simon Rozsival <simon@rozsival.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

@jkotas

Follow-up to #128284 Follow-up to #128385 This PR removes an unnecessary long running unit test added in #128284 /cc @jkotas

Fixes Android X509Store PAL cleanup paths for certificate/private-key entries and JNI local references to avoid memory leaks. Follow-up to #128284 ## Changes - Dispose the Android `KeyStore.PrivateKeyEntry` wrapper held by `AndroidCertificatePal`. - Release JNI local references on Android X509Store cleanup paths: - trusted certificate enumeration - default store open failure/success cleanup - remove-certificate early success path - Add JNI exception checks when advancing Android KeyStore alias enumerations. --------- Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

simonrozsival · 2026-05-16T13:01:15Z

Fixes a JNI global-reference leak in the Android X509 chain PAL.

AndroidCryptoNative_X509ChainGetCertificates returns caller-owned JNI global refs for the certificates in the built chain. Constructing X509Certificate2 from those pointers duplicates the refs for managed ownership, but the original returned refs were never released. This releases those temporary refs after the managed certificates have been created.

Regression test

BuildChainRepeatedly_DoesNotExhaustGlobalReferences is a new [OuterLoop] Android-only test that builds 8,600 6-certificate PKI chains via CertificateAuthority.BuildPrivatePki. Without this PR each successful build leaks 6 JNI global references, so 8,600 iterations would leak 51,600 — past Android's default 51,200 entry limit.

Validation

Run on a local Android emulator (API 36, arm64-v8a, Apple Silicon host, Mono interpreter):

With this PR's fix: ✅ passes (xharness exit 0) in ~9.5 minutes.
Without the fix (verified by reverting the new try/finally block in Interop.X509Chain.cs): ❌ crashes between iter ~7,750 and ~8,500 with JNI ERROR (app bug): global reference table overflow (max=51200) and a tombstone; xharness exit 80 (APP_CRASH).

The Android X509 chain PAL returns JNI global references for chain certificates. Creating X509Certificate2 from those pointers duplicates the references for managed ownership, leaving the native-returned references caller-owned. Release those temporary references after conversion so repeated chain builds do not exhaust ART global refs. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

dotnet-policy-service · 2026-05-16T13:12:05Z

Tagging subscribers to this area: @dotnet/ncl, @bartonjs, @vcsjones
See info in area-owners.md if you want to be subscribed.

Copilot

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.

…ref limit Adds [OuterLoop] regression test BuildChainRepeatedly_DoesNotExhaustGlobalReferences that builds a 6-certificate chain (root + 4 intermediates + endCert) 8,600 times. Without the gref-release fix in PR dotnet#128284, AndroidCryptoNative_X509ChainGetCertificates returns caller-owned JNI global references that X509Certificate2 then duplicates, leaving the native-returned refs orphaned. A 6-cert chain leaks 6 grefs per successful build; 8,600 builds therefore leak 51,600 references, which exceeds Android's default global-reference table limit (51,200) and aborts the process with 'global reference table overflow (max=51200)'. Threshold validation against the unfixed code path: - 8,400 iterations completed successfully - 8,500 iterations crashed with the gref-table overflow With the managed try/finally cleanup in place (Interop.X509Chain.X509ChainGetCertificates), 8,600 iterations complete cleanly. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…ref limit Adds [OuterLoop] companion test BuildChainRepeatedlyOnFailure_DoesNotExhaustGlobalReferences that builds a 6-certificate chain and forces it to fail (VerificationTime far in the future, well outside cert validity) 60,000 times. The success-path test (BuildChainRepeatedly_DoesNotExhaustGlobalReferences) exercises X509ChainGetCertificates, which PR dotnet#128284 fixed. The failure path is a different code path: managed code skips GetCertificates entirely when chain.Build returns false, and instead invokes X509ChainGetErrors and walks the resulting error list. The pal_x509chain.c side allocates xmalloc'd UTF-16 message buffers per error; the managed side is expected to Marshal.FreeHGlobal each one. The failure path doesn't currently leak — the test is defensive against future regressions that introduce a per-iteration leak on this path (e.g., a forgotten Marshal.FreeHGlobal, a missing ReleaseGRef on a throwable, or a regression in X509ChainBuild's exception-handling flow). 60,000 iterations is chosen to overflow the 51,200-entry JNI global reference table even under a worst-case 1-gref-per-iteration regression. The 6-cert chain mirrors the success-path test so the two tests cover the same chain shape. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Copilot

Pull request overview

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

simonrozsival · 2026-05-19T21:13:24Z

/backport to release/10.0

github-actions · 2026-05-19T21:13:34Z

Started backporting to release/10.0 (link to workflow run)

@simonrozsival

Backport of #128284 to release/10.0 /cc @simonrozsival ## Customer Impact - [x] Customer reported - [ ] Found internally A customer reported that their app was crashing due to a memory leak in X509Chain class (JNI global ref table overflow). ## Regression - [ ] Yes - [x] No [If yes, specify when the regression was introduced. Provide the PR or commit if known.] ## Testing The new unit test (ChainTests.BuildChainRepeatedly_DoesNotExhaustGlobalReferences) passes locally on Android emulator. ## Risk Low. The change only affects Android and the code path is well covered by tests. **IMPORTANT**: If this backport is for a servicing release, please verify that: - For .NET 8 and .NET 9: The PR target branch is `release/X.0-staging`, not `release/X.0`. - For .NET 10+: The PR target branch is `release/X.0` (no `-staging` suffix). ## Package authoring no longer needed in .NET 9 **IMPORTANT**: Starting with .NET 9, you no longer need to edit a NuGet package's csproj to enable building and bump the version. Keep in mind that we still need package authoring in .NET 8 and older versions. --------- Co-authored-by: Simon Rozsival <simon@rozsival.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

@jkotas

Follow-up to #128284 Follow-up to #128385 This PR removes an unnecessary long running unit test added in #128284 /cc @jkotas

Fixes Android X509Store PAL cleanup paths for certificate/private-key entries and JNI local references to avoid memory leaks. Follow-up to #128284 ## Changes - Dispose the Android `KeyStore.PrivateKeyEntry` wrapper held by `AndroidCertificatePal`. - Release JNI local references on Android X509Store cleanup paths: - trusted certificate enumeration - default store open failure/success cleanup - remove-certificate early success path - Add JNI exception checks when advancing Android KeyStore alias enumerations. --------- Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

simonrozsival · 2026-05-16T13:01:15Z

Fixes a JNI global-reference leak in the Android X509 chain PAL.

AndroidCryptoNative_X509ChainGetCertificates returns caller-owned JNI global refs for the certificates in the built chain. Constructing X509Certificate2 from those pointers duplicates the refs for managed ownership, but the original returned refs were never released. This releases those temporary refs after the managed certificates have been created.

Regression test

BuildChainRepeatedly_DoesNotExhaustGlobalReferences is a new [OuterLoop] Android-only test that builds 8,600 6-certificate PKI chains via CertificateAuthority.BuildPrivatePki. Without this PR each successful build leaks 6 JNI global references, so 8,600 iterations would leak 51,600 — past Android's default 51,200 entry limit.

Validation

Run on a local Android emulator (API 36, arm64-v8a, Apple Silicon host, Mono interpreter):

With this PR's fix: ✅ passes (xharness exit 0) in ~9.5 minutes.
Without the fix (verified by reverting the new try/finally block in Interop.X509Chain.cs): ❌ crashes between iter ~7,750 and ~8,500 with JNI ERROR (app bug): global reference table overflow (max=51200) and a tombstone; xharness exit 80 (APP_CRASH).

The Android X509 chain PAL returns JNI global references for chain certificates. Creating X509Certificate2 from those pointers duplicates the references for managed ownership, leaving the native-returned references caller-owned. Release those temporary references after conversion so repeated chain builds do not exhaust ART global refs. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

dotnet-policy-service · 2026-05-16T13:12:05Z

Tagging subscribers to this area: @dotnet/ncl, @bartonjs, @vcsjones
See info in area-owners.md if you want to be subscribed.

Copilot

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.

…ref limit Adds [OuterLoop] regression test BuildChainRepeatedly_DoesNotExhaustGlobalReferences that builds a 6-certificate chain (root + 4 intermediates + endCert) 8,600 times. Without the gref-release fix in PR dotnet#128284, AndroidCryptoNative_X509ChainGetCertificates returns caller-owned JNI global references that X509Certificate2 then duplicates, leaving the native-returned refs orphaned. A 6-cert chain leaks 6 grefs per successful build; 8,600 builds therefore leak 51,600 references, which exceeds Android's default global-reference table limit (51,200) and aborts the process with 'global reference table overflow (max=51200)'. Threshold validation against the unfixed code path: - 8,400 iterations completed successfully - 8,500 iterations crashed with the gref-table overflow With the managed try/finally cleanup in place (Interop.X509Chain.X509ChainGetCertificates), 8,600 iterations complete cleanly. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…ref limit Adds [OuterLoop] companion test BuildChainRepeatedlyOnFailure_DoesNotExhaustGlobalReferences that builds a 6-certificate chain and forces it to fail (VerificationTime far in the future, well outside cert validity) 60,000 times. The success-path test (BuildChainRepeatedly_DoesNotExhaustGlobalReferences) exercises X509ChainGetCertificates, which PR dotnet#128284 fixed. The failure path is a different code path: managed code skips GetCertificates entirely when chain.Build returns false, and instead invokes X509ChainGetErrors and walks the resulting error list. The pal_x509chain.c side allocates xmalloc'd UTF-16 message buffers per error; the managed side is expected to Marshal.FreeHGlobal each one. The failure path doesn't currently leak — the test is defensive against future regressions that introduce a per-iteration leak on this path (e.g., a forgotten Marshal.FreeHGlobal, a missing ReleaseGRef on a throwable, or a regression in X509ChainBuild's exception-handling flow). 60,000 iterations is chosen to overflow the 51,200-entry JNI global reference table even under a worst-case 1-gref-per-iteration regression. The 6-cert chain mirrors the success-path test so the two tests cover the same chain shape. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Copilot

Pull request overview

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

simonrozsival · 2026-05-19T21:13:24Z

/backport to release/10.0

github-actions · 2026-05-19T21:13:34Z

Started backporting to release/10.0 (link to workflow run)

@simonrozsival

Backport of #128284 to release/10.0 /cc @simonrozsival ## Customer Impact - [x] Customer reported - [ ] Found internally A customer reported that their app was crashing due to a memory leak in X509Chain class (JNI global ref table overflow). ## Regression - [ ] Yes - [x] No [If yes, specify when the regression was introduced. Provide the PR or commit if known.] ## Testing The new unit test (ChainTests.BuildChainRepeatedly_DoesNotExhaustGlobalReferences) passes locally on Android emulator. ## Risk Low. The change only affects Android and the code path is well covered by tests. **IMPORTANT**: If this backport is for a servicing release, please verify that: - For .NET 8 and .NET 9: The PR target branch is `release/X.0-staging`, not `release/X.0`. - For .NET 10+: The PR target branch is `release/X.0` (no `-staging` suffix). ## Package authoring no longer needed in .NET 9 **IMPORTANT**: Starting with .NET 9, you no longer need to edit a NuGet package's csproj to enable building and bump the version. Keep in mind that we still need package authoring in .NET 8 and older versions. --------- Co-authored-by: Simon Rozsival <simon@rozsival.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

@jkotas

Follow-up to #128284 Follow-up to #128385 This PR removes an unnecessary long running unit test added in #128284 /cc @jkotas

Fixes Android X509Store PAL cleanup paths for certificate/private-key entries and JNI local references to avoid memory leaks. Follow-up to #128284 ## Changes - Dispose the Android `KeyStore.PrivateKeyEntry` wrapper held by `AndroidCertificatePal`. - Release JNI local references on Android X509Store cleanup paths: - trusted certificate enumeration - default store open failure/success cleanup - remove-certificate early success path - Add JNI exception checks when advancing Android KeyStore alias enumerations. --------- Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

simonrozsival · 2026-05-16T13:01:15Z

Fixes a JNI global-reference leak in the Android X509 chain PAL.

AndroidCryptoNative_X509ChainGetCertificates returns caller-owned JNI global refs for the certificates in the built chain. Constructing X509Certificate2 from those pointers duplicates the refs for managed ownership, but the original returned refs were never released. This releases those temporary refs after the managed certificates have been created.

Regression test

BuildChainRepeatedly_DoesNotExhaustGlobalReferences is a new [OuterLoop] Android-only test that builds 8,600 6-certificate PKI chains via CertificateAuthority.BuildPrivatePki. Without this PR each successful build leaks 6 JNI global references, so 8,600 iterations would leak 51,600 — past Android's default 51,200 entry limit.

Validation

Run on a local Android emulator (API 36, arm64-v8a, Apple Silicon host, Mono interpreter):

With this PR's fix: ✅ passes (xharness exit 0) in ~9.5 minutes.
Without the fix (verified by reverting the new try/finally block in Interop.X509Chain.cs): ❌ crashes between iter ~7,750 and ~8,500 with JNI ERROR (app bug): global reference table overflow (max=51200) and a tombstone; xharness exit 80 (APP_CRASH).

The Android X509 chain PAL returns JNI global references for chain certificates. Creating X509Certificate2 from those pointers duplicates the references for managed ownership, leaving the native-returned references caller-owned. Release those temporary references after conversion so repeated chain builds do not exhaust ART global refs. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

dotnet-policy-service · 2026-05-16T13:12:05Z

Tagging subscribers to this area: @dotnet/ncl, @bartonjs, @vcsjones
See info in area-owners.md if you want to be subscribed.

Copilot

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.

…ref limit Adds [OuterLoop] regression test BuildChainRepeatedly_DoesNotExhaustGlobalReferences that builds a 6-certificate chain (root + 4 intermediates + endCert) 8,600 times. Without the gref-release fix in PR dotnet#128284, AndroidCryptoNative_X509ChainGetCertificates returns caller-owned JNI global references that X509Certificate2 then duplicates, leaving the native-returned refs orphaned. A 6-cert chain leaks 6 grefs per successful build; 8,600 builds therefore leak 51,600 references, which exceeds Android's default global-reference table limit (51,200) and aborts the process with 'global reference table overflow (max=51200)'. Threshold validation against the unfixed code path: - 8,400 iterations completed successfully - 8,500 iterations crashed with the gref-table overflow With the managed try/finally cleanup in place (Interop.X509Chain.X509ChainGetCertificates), 8,600 iterations complete cleanly. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…ref limit Adds [OuterLoop] companion test BuildChainRepeatedlyOnFailure_DoesNotExhaustGlobalReferences that builds a 6-certificate chain and forces it to fail (VerificationTime far in the future, well outside cert validity) 60,000 times. The success-path test (BuildChainRepeatedly_DoesNotExhaustGlobalReferences) exercises X509ChainGetCertificates, which PR dotnet#128284 fixed. The failure path is a different code path: managed code skips GetCertificates entirely when chain.Build returns false, and instead invokes X509ChainGetErrors and walks the resulting error list. The pal_x509chain.c side allocates xmalloc'd UTF-16 message buffers per error; the managed side is expected to Marshal.FreeHGlobal each one. The failure path doesn't currently leak — the test is defensive against future regressions that introduce a per-iteration leak on this path (e.g., a forgotten Marshal.FreeHGlobal, a missing ReleaseGRef on a throwable, or a regression in X509ChainBuild's exception-handling flow). 60,000 iterations is chosen to overflow the 51,200-entry JNI global reference table even under a worst-case 1-gref-per-iteration regression. The 6-cert chain mirrors the success-path test so the two tests cover the same chain shape. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Copilot

Pull request overview

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

simonrozsival · 2026-05-19T21:13:24Z

/backport to release/10.0

github-actions · 2026-05-19T21:13:34Z

Started backporting to release/10.0 (link to workflow run)

@simonrozsival

Backport of #128284 to release/10.0 /cc @simonrozsival ## Customer Impact - [x] Customer reported - [ ] Found internally A customer reported that their app was crashing due to a memory leak in X509Chain class (JNI global ref table overflow). ## Regression - [ ] Yes - [x] No [If yes, specify when the regression was introduced. Provide the PR or commit if known.] ## Testing The new unit test (ChainTests.BuildChainRepeatedly_DoesNotExhaustGlobalReferences) passes locally on Android emulator. ## Risk Low. The change only affects Android and the code path is well covered by tests. **IMPORTANT**: If this backport is for a servicing release, please verify that: - For .NET 8 and .NET 9: The PR target branch is `release/X.0-staging`, not `release/X.0`. - For .NET 10+: The PR target branch is `release/X.0` (no `-staging` suffix). ## Package authoring no longer needed in .NET 9 **IMPORTANT**: Starting with .NET 9, you no longer need to edit a NuGet package's csproj to enable building and bump the version. Keep in mind that we still need package authoring in .NET 8 and older versions. --------- Co-authored-by: Simon Rozsival <simon@rozsival.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

@jkotas

Follow-up to #128284 Follow-up to #128385 This PR removes an unnecessary long running unit test added in #128284 /cc @jkotas

Fixes Android X509Store PAL cleanup paths for certificate/private-key entries and JNI local references to avoid memory leaks. Follow-up to #128284 ## Changes - Dispose the Android `KeyStore.PrivateKeyEntry` wrapper held by `AndroidCertificatePal`. - Release JNI local references on Android X509Store cleanup paths: - trusted certificate enumeration - default store open failure/success cleanup - remove-certificate early success path - Add JNI exception checks when advancing Android KeyStore alias enumerations. --------- Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

simonrozsival · 2026-05-16T13:01:15Z

Fixes a JNI global-reference leak in the Android X509 chain PAL.

AndroidCryptoNative_X509ChainGetCertificates returns caller-owned JNI global refs for the certificates in the built chain. Constructing X509Certificate2 from those pointers duplicates the refs for managed ownership, but the original returned refs were never released. This releases those temporary refs after the managed certificates have been created.

Regression test

BuildChainRepeatedly_DoesNotExhaustGlobalReferences is a new [OuterLoop] Android-only test that builds 8,600 6-certificate PKI chains via CertificateAuthority.BuildPrivatePki. Without this PR each successful build leaks 6 JNI global references, so 8,600 iterations would leak 51,600 — past Android's default 51,200 entry limit.

Validation

Run on a local Android emulator (API 36, arm64-v8a, Apple Silicon host, Mono interpreter):

With this PR's fix: ✅ passes (xharness exit 0) in ~9.5 minutes.
Without the fix (verified by reverting the new try/finally block in Interop.X509Chain.cs): ❌ crashes between iter ~7,750 and ~8,500 with JNI ERROR (app bug): global reference table overflow (max=51200) and a tombstone; xharness exit 80 (APP_CRASH).

The Android X509 chain PAL returns JNI global references for chain certificates. Creating X509Certificate2 from those pointers duplicates the references for managed ownership, leaving the native-returned references caller-owned. Release those temporary references after conversion so repeated chain builds do not exhaust ART global refs. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

dotnet-policy-service · 2026-05-16T13:12:05Z

Tagging subscribers to this area: @dotnet/ncl, @bartonjs, @vcsjones
See info in area-owners.md if you want to be subscribed.

Copilot

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.

…ref limit Adds [OuterLoop] regression test BuildChainRepeatedly_DoesNotExhaustGlobalReferences that builds a 6-certificate chain (root + 4 intermediates + endCert) 8,600 times. Without the gref-release fix in PR dotnet#128284, AndroidCryptoNative_X509ChainGetCertificates returns caller-owned JNI global references that X509Certificate2 then duplicates, leaving the native-returned refs orphaned. A 6-cert chain leaks 6 grefs per successful build; 8,600 builds therefore leak 51,600 references, which exceeds Android's default global-reference table limit (51,200) and aborts the process with 'global reference table overflow (max=51200)'. Threshold validation against the unfixed code path: - 8,400 iterations completed successfully - 8,500 iterations crashed with the gref-table overflow With the managed try/finally cleanup in place (Interop.X509Chain.X509ChainGetCertificates), 8,600 iterations complete cleanly. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…ref limit Adds [OuterLoop] companion test BuildChainRepeatedlyOnFailure_DoesNotExhaustGlobalReferences that builds a 6-certificate chain and forces it to fail (VerificationTime far in the future, well outside cert validity) 60,000 times. The success-path test (BuildChainRepeatedly_DoesNotExhaustGlobalReferences) exercises X509ChainGetCertificates, which PR dotnet#128284 fixed. The failure path is a different code path: managed code skips GetCertificates entirely when chain.Build returns false, and instead invokes X509ChainGetErrors and walks the resulting error list. The pal_x509chain.c side allocates xmalloc'd UTF-16 message buffers per error; the managed side is expected to Marshal.FreeHGlobal each one. The failure path doesn't currently leak — the test is defensive against future regressions that introduce a per-iteration leak on this path (e.g., a forgotten Marshal.FreeHGlobal, a missing ReleaseGRef on a throwable, or a regression in X509ChainBuild's exception-handling flow). 60,000 iterations is chosen to overflow the 51,200-entry JNI global reference table even under a worst-case 1-gref-per-iteration regression. The 6-cert chain mirrors the success-path test so the two tests cover the same chain shape. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Copilot

Pull request overview

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

simonrozsival · 2026-05-19T21:13:24Z

/backport to release/10.0

github-actions · 2026-05-19T21:13:34Z

Started backporting to release/10.0 (link to workflow run)

@simonrozsival

Backport of #128284 to release/10.0 /cc @simonrozsival ## Customer Impact - [x] Customer reported - [ ] Found internally A customer reported that their app was crashing due to a memory leak in X509Chain class (JNI global ref table overflow). ## Regression - [ ] Yes - [x] No [If yes, specify when the regression was introduced. Provide the PR or commit if known.] ## Testing The new unit test (ChainTests.BuildChainRepeatedly_DoesNotExhaustGlobalReferences) passes locally on Android emulator. ## Risk Low. The change only affects Android and the code path is well covered by tests. **IMPORTANT**: If this backport is for a servicing release, please verify that: - For .NET 8 and .NET 9: The PR target branch is `release/X.0-staging`, not `release/X.0`. - For .NET 10+: The PR target branch is `release/X.0` (no `-staging` suffix). ## Package authoring no longer needed in .NET 9 **IMPORTANT**: Starting with .NET 9, you no longer need to edit a NuGet package's csproj to enable building and bump the version. Keep in mind that we still need package authoring in .NET 8 and older versions. --------- Co-authored-by: Simon Rozsival <simon@rozsival.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

@jkotas

Follow-up to #128284 Follow-up to #128385 This PR removes an unnecessary long running unit test added in #128284 /cc @jkotas

Fixes Android X509Store PAL cleanup paths for certificate/private-key entries and JNI local references to avoid memory leaks. Follow-up to #128284 ## Changes - Dispose the Android `KeyStore.PrivateKeyEntry` wrapper held by `AndroidCertificatePal`. - Release JNI local references on Android X509Store cleanup paths: - trusted certificate enumeration - default store open failure/success cleanup - remove-certificate early success path - Add JNI exception checks when advancing Android KeyStore alias enumerations. --------- Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

simonrozsival · 2026-05-16T13:01:15Z

Fixes a JNI global-reference leak in the Android X509 chain PAL.

AndroidCryptoNative_X509ChainGetCertificates returns caller-owned JNI global refs for the certificates in the built chain. Constructing X509Certificate2 from those pointers duplicates the refs for managed ownership, but the original returned refs were never released. This releases those temporary refs after the managed certificates have been created.

Regression test

BuildChainRepeatedly_DoesNotExhaustGlobalReferences is a new [OuterLoop] Android-only test that builds 8,600 6-certificate PKI chains via CertificateAuthority.BuildPrivatePki. Without this PR each successful build leaks 6 JNI global references, so 8,600 iterations would leak 51,600 — past Android's default 51,200 entry limit.

Validation

Run on a local Android emulator (API 36, arm64-v8a, Apple Silicon host, Mono interpreter):

With this PR's fix: ✅ passes (xharness exit 0) in ~9.5 minutes.
Without the fix (verified by reverting the new try/finally block in Interop.X509Chain.cs): ❌ crashes between iter ~7,750 and ~8,500 with JNI ERROR (app bug): global reference table overflow (max=51200) and a tombstone; xharness exit 80 (APP_CRASH).

The Android X509 chain PAL returns JNI global references for chain certificates. Creating X509Certificate2 from those pointers duplicates the references for managed ownership, leaving the native-returned references caller-owned. Release those temporary references after conversion so repeated chain builds do not exhaust ART global refs. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

dotnet-policy-service · 2026-05-16T13:12:05Z

Tagging subscribers to this area: @dotnet/ncl, @bartonjs, @vcsjones
See info in area-owners.md if you want to be subscribed.

Copilot

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.

…ref limit Adds [OuterLoop] regression test BuildChainRepeatedly_DoesNotExhaustGlobalReferences that builds a 6-certificate chain (root + 4 intermediates + endCert) 8,600 times. Without the gref-release fix in PR dotnet#128284, AndroidCryptoNative_X509ChainGetCertificates returns caller-owned JNI global references that X509Certificate2 then duplicates, leaving the native-returned refs orphaned. A 6-cert chain leaks 6 grefs per successful build; 8,600 builds therefore leak 51,600 references, which exceeds Android's default global-reference table limit (51,200) and aborts the process with 'global reference table overflow (max=51200)'. Threshold validation against the unfixed code path: - 8,400 iterations completed successfully - 8,500 iterations crashed with the gref-table overflow With the managed try/finally cleanup in place (Interop.X509Chain.X509ChainGetCertificates), 8,600 iterations complete cleanly. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…ref limit Adds [OuterLoop] companion test BuildChainRepeatedlyOnFailure_DoesNotExhaustGlobalReferences that builds a 6-certificate chain and forces it to fail (VerificationTime far in the future, well outside cert validity) 60,000 times. The success-path test (BuildChainRepeatedly_DoesNotExhaustGlobalReferences) exercises X509ChainGetCertificates, which PR dotnet#128284 fixed. The failure path is a different code path: managed code skips GetCertificates entirely when chain.Build returns false, and instead invokes X509ChainGetErrors and walks the resulting error list. The pal_x509chain.c side allocates xmalloc'd UTF-16 message buffers per error; the managed side is expected to Marshal.FreeHGlobal each one. The failure path doesn't currently leak — the test is defensive against future regressions that introduce a per-iteration leak on this path (e.g., a forgotten Marshal.FreeHGlobal, a missing ReleaseGRef on a throwable, or a regression in X509ChainBuild's exception-handling flow). 60,000 iterations is chosen to overflow the 51,200-entry JNI global reference table even under a worst-case 1-gref-per-iteration regression. The 6-cert chain mirrors the success-path test so the two tests cover the same chain shape. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Copilot

Pull request overview

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

simonrozsival · 2026-05-19T21:13:24Z

/backport to release/10.0

github-actions · 2026-05-19T21:13:34Z

Started backporting to release/10.0 (link to workflow run)

@simonrozsival

Backport of #128284 to release/10.0 /cc @simonrozsival ## Customer Impact - [x] Customer reported - [ ] Found internally A customer reported that their app was crashing due to a memory leak in X509Chain class (JNI global ref table overflow). ## Regression - [ ] Yes - [x] No [If yes, specify when the regression was introduced. Provide the PR or commit if known.] ## Testing The new unit test (ChainTests.BuildChainRepeatedly_DoesNotExhaustGlobalReferences) passes locally on Android emulator. ## Risk Low. The change only affects Android and the code path is well covered by tests. **IMPORTANT**: If this backport is for a servicing release, please verify that: - For .NET 8 and .NET 9: The PR target branch is `release/X.0-staging`, not `release/X.0`. - For .NET 10+: The PR target branch is `release/X.0` (no `-staging` suffix). ## Package authoring no longer needed in .NET 9 **IMPORTANT**: Starting with .NET 9, you no longer need to edit a NuGet package's csproj to enable building and bump the version. Keep in mind that we still need package authoring in .NET 8 and older versions. --------- Co-authored-by: Simon Rozsival <simon@rozsival.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

@jkotas

Follow-up to #128284 Follow-up to #128385 This PR removes an unnecessary long running unit test added in #128284 /cc @jkotas

Fixes Android X509Store PAL cleanup paths for certificate/private-key entries and JNI local references to avoid memory leaks. Follow-up to #128284 ## Changes - Dispose the Android `KeyStore.PrivateKeyEntry` wrapper held by `AndroidCertificatePal`. - Release JNI local references on Android X509Store cleanup paths: - trusted certificate enumeration - default store open failure/success cleanup - remove-certificate early success path - Add JNI exception checks when advancing Android KeyStore alias enumerations. --------- Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

CopilotAI review requested due to automatic review settings May 16, 2026 13:01

simonrozsival requested review from MichalStrehovsky, steveisok and vitek-karas as code owners May 16, 2026 13:01

Copilotstarted reviewing on behalf of simonrozsival May 16, 2026 13:01 View session

github-actionsBot added the area-crossgen2-coreclr only use for closed issues label May 16, 2026

dotnet-policy-serviceBot assigned simonrozsival May 16, 2026

simonrozsival marked this pull request as draft May 16, 2026 13:03

simonrozsival closed this May 16, 2026

simonrozsival force-pushed the dev/simonrozsival/android-x509-gref-leak-fix branch from 5853fff to cce39fdCompare May 16, 2026 13:05

simonrozsival reopened this May 16, 2026

simonrozsival force-pushed the dev/simonrozsival/android-x509-gref-leak-fix branch from 5853fff to cce39fdCompare May 16, 2026 13:10

simonrozsival removed the request for review from MichalStrehovsky May 16, 2026 13:10

simonrozsival added area-System.Net.Security os-android and removed area-crossgen2-coreclr only use for closed issues labels May 16, 2026

simonrozsival requested a review from vcsjones May 16, 2026 13:11

CopilotAI reviewed May 16, 2026

This was referenced May 16, 2026
Unable to pull image from mcr.microsoft.com #117164
Closed
Multiple Helix work items fail on maccatalyst/tvos CoreCLR Release #126460
Closed

bartonjs approved these changes May 16, 2026
View reviewed changes

vcsjones reviewed May 17, 2026
View reviewed changes

Comment thread.../Common/src/Interop/Android/System.Security.Cryptography.Native.Android/Interop.X509Chain.cs

CopilotAI review requested due to automatic review settings May 18, 2026 13:05

simonrozsival force-pushed the dev/simonrozsival/android-x509-gref-leak-fix branch from 6b38047 to c36b4e7Compare May 18, 2026 13:05

Copilotstarted reviewing on behalf of simonrozsival May 18, 2026 13:06 View session

CopilotAI reviewed May 18, 2026
View reviewed changes

Comment thread.../Common/src/Interop/Android/System.Security.Cryptography.Native.Android/Interop.X509Chain.cs

simonrozsival merged commit 11b17d7 into dotnet:mainMay 19, 2026
89 of 93 checks passed

simonrozsival deleted the dev/simonrozsival/android-x509-gref-leak-fix branch May 19, 2026 21:10

github-actionsBot mentioned this pull request May 19, 2026
[release/10.0] Release Android X509 chain certificate GREFs #128385
Merged
4 tasks

dotnet-maestroBot mentioned this pull request May 20, 2026
[main] Source code updates from dotnet/runtime dotnet/dotnet#6708
Merged

CopilotAI mentioned this pull request May 20, 2026
Relax localhost subdomain address-family checks on Android/AppleMobile #128404
Closed

dotnet-milestone-botBot added this to the 11.0-preview6 milestone May 21, 2026

simonrozsival mentioned this pull request May 22, 2026
[Android] Remove unnecessary X509Chain memory leak test #128493
Merged

jkotas pushed a commit that referenced this pull request May 22, 2026
[Android] Remove unnecessary X509Chain memory leak test (#128493)
75cd6d8
Follow-up to #128284 Follow-up to #128385 This PR removes an unnecessary long running unit test added in #128284 /cc @jkotas

simonrozsival mentioned this pull request May 27, 2026
[Android] Fix X509Store cleanup #128629
Merged

github-actionsBot locked and limited conversation to collaborators Jun 21, 2026