Weasel 9.48.0

Directory.Packages.props

@@ -1118,9 +1118,57 @@
          those same rows and was also unread, but PrimaryKeyName falls back to Weasel's derived
          pkey_ default, which is what Marten declares, so that half was masked here; it bites a
          configuration that names the constraint itself. Nothing public added, removed or changed. -->
-    <PackageVersion Include="Weasel.EntityFrameworkCore" Version="9.47.1" />
-    <PackageVersion Include="Weasel.Postgresql" Version="9.47.1" />
-    <PackageVersion Include="Weasel.Storage" Version="9.47.1" />
+    <!-- Weasel 9.48.0 (from 9.47.1). ADOPTED as-is, no Marten code changes. Three of the six issues
+
+         The JasperFx floor: Weasel.Core 9.48.0 declares JasperFx >= 2.98.0 and Weasel.Storage
+         declares JasperFx.Events >= 2.98.0, both read from the published nuspec. Already satisfied
+         here, because the JasperFx 2.98.0 bump landed on its own ahead of this one; at 2.97.0 the
+         restore fails outright with NU1605, a package downgrade promoted to an error, rather than
+         resolving quietly upward.
+         are SQL Server only and inert here; what reaches Marten is the PostgreSQL parameter-type work
+         and one PostgreSQL partitioning fix.
+
+         weasel#758/#759: the parameter-type catch-all. A private helper in five providers ended in
+         `?? <catch-all>`, which can never return null, so determineParameterType always reported
+         success and every fall-through branch after the first was dead code. PostgreSQL was one of
+         the two providers that already fell through correctly, which is exactly why the divergence
+         read as a Postgres quirk rather than a bug in the other five, so the resolution behaviour
+         Marten sees is unchanged. What does change for Marten is the shared base: ToParameterType now
+         THROWS for a type Npgsql has no mapping for, where some providers used to answer with a
+         character type. RuntimeCompiledQueryDescriptorFactory and PostgresStorageDialect call it, and
+         both pass types Npgsql maps. StoreOptions.Identity and ExecuteSqlStorageOperation use
+         TryGetDbType, which still returns null for an unmapped type. Also fixed in the same PR, and
+         the one piece that was a hard failure rather than a slow query: both AddParameters overloads
+         resolved through ToParameterType on the TYPE, going around the value-aware seam weasel#403
+         added for Npgsql, so a UTC DateTime in a parameter bag was bound as `timestamp without time
+         zone` and Npgsql threw outright, while the same value through AddNamedParameter worked. They
+         now resolve from the value and keep the declared type for a null.
+
+         weasel#764: PostgreSQL's numeric arm in PartitionExtensions.FormatSqlValue rendered with the
+         CURRENT CULTURE, so a decimal partition bound carried a comma separator under de-DE and a
+         negative one a U+2212 minus sign under sv-SE, fi-FI, lt-LT or nb-NO. The date arms were
+         already invariant with a comment saying why, which is what made the numeric one read as
+         deliberate. Marten's tenant and list partitions are string bounds, which take the quoted
+         path, so this bites a numeric partition bound only. The SQL Server half of the issue is where
+         it was found (Polecat, under a th-TH culture).
+
+         weasel#761: an enum parameter resolves by EnumStorage now, with
+         ToParameterType/TryGetDbType/TryGetDbTypeForValue overloads taking one, and PostgreSQL keeps
+         Integer for an int-backed enum exactly as before. The one difference here: an enum now
+         resolves through its UNDERLYING type, so a long-backed enum gives Bigint rather than silently
+         narrowing to Integer, and a byte-backed one Smallint. Arrays are untouched on PostgreSQL,
+         which maps them properly as an Array composite; the four providers that bound an int[] as
+         text now defer instead.
+
+         INERT: weasel#762 (SQL Server string Unicode precedence, documentation only), weasel#763
+         (DateOnly/TimeOnly on MySQL and Oracle), weasel#765 and weasel#766 (SQL Server partition DDL
+         and sp_getapplock).
+
+         Riding along: MySqlConnector 2.6.1 for GHSA-473q-m89c-ghf8, which is not a Weasel.Postgresql
+         surface, and JasperFx / JasperFx.Events 2.98.0, which the pin above satisfies. -->
+    <PackageVersion Include="Weasel.EntityFrameworkCore" Version="9.48.0" />
+    <PackageVersion Include="Weasel.Postgresql" Version="9.48.0" />
+    <PackageVersion Include="Weasel.Storage" Version="9.48.0" />
     <PackageVersion Include="WolverineFx.Marten" Version="4.2.0" />
     <!-- The whole test suite is on xunit v3. VSTest stays the execution mode, so the
          Nuke targets and the GitHub workflows are unaffected. -->