Epic: 10.1.0 #296

Goal

10.1.0 is the first release on the 10.x LTS line after 10.0.0, shipped from release/10.x. It is a minor, not a patch, because it carries one silent fix (#300) and the policy ships those only in a minor that opens with ### Changed answers (CONTRIBUTING.md, "A fix that changes an answer"; decided 2026-10-19). It fixes what consumers of the released 10.0.0 meet on the EF Core 10 patches and the .NET 11 toolchain they actually get — versions above the floor this repository's CI pins — and makes the line's CI see those versions, so the next upstream servicing change is found by CI rather than by chance.

In scope

Fixes, in the order they reach consumers. Each loud one is on main in PR #292, which is a single commit holding the whole .NET 11 retarget, so each port is a hand-extracted commit naming the PR — or #292 is split before it merges so cherry-pick -x works.

• 10.x: Any/Count over an empty captured list of ranges or value sets fails with 42883: operator does not exist: text && int4range on EF Core 10.0.2 and later (dotnet/efcore#37216) #302 Any/Count over an empty captured list of ranges or value sets fails on the server (42883: operator does not exist: text && int4range) on EF Core 10.0.2 and later (EF Core 10: InvalidOperationException doing primitive list join when list is empty dotnet/efcore#37216). Live today: Npgsql EF 10.0.1–10.0.3 require EF Core ≥ 10.0.4. Measured 2026-10-19 on EF Core 10.0.13 / Npgsql EF 10.0.3 against the 10.0.0 code: 4 translation and 5 integration tests fail; AppContext.SetSwitch("Microsoft.EntityFrameworkCore.Issue37216", true) is the measured stopgap. The port keeps the 10.0.0 path for the floor.
• 10.x: RangeSet.Contains throws ArgumentException on the EF Core 10 patch that carries dotnet/efcore#37081 #288 RangeSet.Contains throws ArgumentException on the EF Core 10 patch that carries [release/10.0] Fix Contains handling for read-only collections dotnet/efcore#39114 — milestone 10.0.15, so pending; stopgap Microsoft.EntityFrameworkCore.Issue37081. Reported upstream as Collection Contains normalization rewrites any Contains overload and guesses the element type from the first type argument dotnet/efcore#39329.
• 10.x: Native AOT published with the .NET 11 SDK traps (SIGTRAP) on the first value set ToString (dotnet/runtime#136073) #299 Native AOT published with the .NET 11 SDK traps (SIGTRAP) on the first value set ToString (NativeAOT (.NET 11 RC2, osx-arm64): SIGTRAP in the instantiating stub of a default interface IFormattable.ToString on a shared-generic interface dotnet/runtime#136073). Measured against the released 10.0.0 package; the workaround is surface-neutral.
• 10.x: on Npgsql EF 10.0.3 with PostgreSQL 16 configured, the JSON-setter refusal names the column instead of the path, and gives a reason that no longer holds (jsonb_set_lax) #303 on Npgsql EF 10.0.3 with SetPostgresVersion(16, 0) or higher, the JSON-setter refusal names the column instead of the property path and gives a reason jsonb_set_lax makes false (Fix JSON null partial updates npgsql/efcore.pg#3855). Still refused in every case; cosmetic, severity/low. Measured: 6 of 7 ExecuteUpdateGuardTests fail with the version configured and all pass with it unset (Npgsql EF 10 assumes 14), which is why the suites did not see it.
• 10.x on the .NET 11 runtime reads an hour of 24: {24:00:00} parses as {00:00:00} and T24:00:00 as the next day #300 the released 10.0.0 on the .NET 11 runtime reads an hour of 24 as midnight: {24:00:00} parses as {00:00:00} and T24:00:00 as the next day, where .NET 10 throws. silent, severity/medium, the reason this is a minor. Its PR adds the ### Changed answers row at the top of the ## [10.1.0] section and the ## Changed answers in 10.1.0 section of docs/migration.md; pr-hygiene.sh refuses a silent PR without the row and reads the version from the branch, so it comes after Release 10.1.0: open the version on release/10.x, date the changelogs and tag v10.1.0 #301 step 1.

CI

• 10.x CI: run the EF Core suites on the newest EF Core / Npgsql 10 patch and the daily servicing builds, not only the floor #297 the EF suites on release/10.x also run on the newest EF Core / Npgsql 10 patch (PR test: run the EF suites on the newest EF Core / Npgsql EF 10 patch as well as the floor #311). The run against the daily servicing builds moved to 10.x CI: run the EF suites weekly against EF Core's staged 10.0 servicing build (darc-pub feed), scheduled from main #312 in the backlog on 2026-10-19: the dotnet10 feed holds no servicing build, and schedule fires only from main. The one-off run that found 10.x: Any/Count over an empty captured list of ranges or value sets fails with 42883: operator does not exist: text && int4range on EF Core 10.0.2 and later (dotnet/efcore#37216) #302 is the mechanism to start from (described on the issue); a .NET 11 runtime roll-forward leg would catch 10.x on the .NET 11 runtime reads an hour of 24: {24:00:00} parses as {00:00:00} and T24:00:00 as the next day #300.

Release

• Release 10.1.0: open the version on release/10.x, date the changelogs and tag v10.1.0 #301 open 10.1.0 on release/10.x (version, ApiCompat baseline, SECURITY.md row, changelog sections), then date the changelogs and tag v10.1.0.

Sequence

1. Release 10.1.0: open the version on release/10.x, date the changelogs and tag v10.1.0 #301 step 1 (open the version) merges first, so every fix PR adds its line under ## [10.1.0] and the hygiene check sees a minor.
2. build: retarget main to .NET 11 for 11.0.0-preview.1 #292 merges into main — every fix exists there first (CONTRIBUTING.md) — then one PR per fix against release/10.x: 10.x: Any/Count over an empty captured list of ranges or value sets fails with 42883: operator does not exist: text && int4range on EF Core 10.0.2 and later (dotnet/efcore#37216) #302, 10.x: RangeSet.Contains throws ArgumentException on the EF Core 10 patch that carries dotnet/efcore#37081 #288, 10.x: Native AOT published with the .NET 11 SDK traps (SIGTRAP) on the first value set ToString (dotnet/runtime#136073) #299, 10.x: on Npgsql EF 10.0.3 with PostgreSQL 16 configured, the JSON-setter refusal names the column instead of the path, and gives a reason that no longer holds (jsonb_set_lax) #303, 10.x on the .NET 11 runtime reads an hour of 24: {24:00:00} parses as {00:00:00} and T24:00:00 as the next day #300. Each proves itself by reverting: its tests fail on the newest patch (for 10.x: Native AOT published with the .NET 11 SDK traps (SIGTRAP) on the first value set ToString (dotnet/runtime#136073) #299 under the .NET 11 SDK, for 10.x on the .NET 11 runtime reads an hour of 24: {24:00:00} parses as {00:00:00} and T24:00:00 as the next day #300 on the .NET 11 runtime), the floor stays green.
3. 10.x CI: run the EF Core suites on the newest EF Core / Npgsql 10 patch and the daily servicing builds, not only the floor #297 lands so the newest-patch leg guards the fixes from then on. Its done-when and 10.x: RangeSet.Contains throws ArgumentException on the EF Core 10 patch that carries dotnet/efcore#37081 #288's are measurable today; the re-check against EF Core 10.0.15 is a comment when that patch ships, never a gate.
4. Release 10.1.0: open the version on release/10.x, date the changelogs and tag v10.1.0 #301 steps 3–4: date, tag, close the 10.x-only issues by hand, add the 10.1.0 section to main's changelogs.

Not in scope

• Registry of EF Core servicing off switches (IssueNNNN) between the floor and the newest 10.0 patch, each with a verdict #298 (the registry of EF Core off switches with a verdict each): classifies upstream changes without changing the packages, so it lives under the oracle theme (Epic: Systematic bug finding with generated oracles #36). Its two verdicts so far point here (Issue37216 → 10.x: Any/Count over an empty captured list of ranges or value sets fails with 42883: operator does not exist: text && int4range on EF Core 10.0.2 and later (dotnet/efcore#37216) #302, Issue37081 → 10.x: RangeSet.Contains throws ArgumentException on the EF Core 10 patch that carries dotnet/efcore#37081 #288).
• Native AOT compiled models (dbcontext optimize --nativeaot) fail: plugin type mappings have no public static Default #209 (Native AOT compiled models): needs a public Default on the mappings, an addition with its own design; a later 10.x minor.
• Inventory where public types meet framework name/interface heuristics, and sweep every type × heuristic pair #293, Every EF Core 11 behaviour-change test records whether the upstream change was backported to 10.0 #294, Sweep the EF translators for name-only dispatch and default operand classification #295: the oracle theme (Epic: Systematic bug finding with generated oracles #36), not bound to a release.
• Additions beyond the fixes (a 10.x minor may carry them inside the window CONTRIBUTING.md sets, but none is planned here), and anything breaking.

Release

Tag v10.1.0 on release/10.x. The root changelog section opens with ### Changed answers (one row, #300) before the fixes. Issues fixed only there are closed by hand (Closes #n acts only on main).