Fixed an issue causing exit tests to return invalid results when no standard output or error is produced (#1695)
Emits a warning when one of the testing library’s attribute macros, such as @Test, includes generic type parameters since they are unsupported. (#1653)
libdispatch
Changed dispatch worker thread names to “DispatchWorker”, instead of being based on the label of the root queue they are servicing (#936)
Swift Build
When using resources, allow the Bundle.module to be accessed from outside the MainActor (#1240)
SourceKit-LSP
Fixed an issue that could cause sourcekit-lsp to crash on Windows when using a compile_commands.json for build settings (#2606)
Introduced forceResolvedVersions configuration option, which is equivalent to SwiftPM’s --force-resolved-versionsoption. This can be used to ensure that background indexing treats Package.resolved as a lock file (#2613)
You should use the --tag swift-6.3.2-RELEASE option. Additionally, Swift 6.3.2 was released from the release/6.3 branch, so you can also use that scheme too.
I`m not sure: are all these changes in Swift 6.3.2 really - I mean really really - BugFixes? It looks like somethink are more than patches and its feels wrong to provide this in a patch version 6.3.2 instead of 6.4 - or not?
Swift doesn't strictly follow semantic versioning. Though personally I'd consider them all fixes of one type or another. SourceKit-LSP's forceResolvedVersions is as close to a new feature I can see, but even that is just "fixing" its mismatched behavior vs. SPM.
Nice, have you tried running the compiler validation suite when natively building on FreeBSD AArch64, by passing the -T flag to build-script as I showed here for linux AArch64? It would be great if you filed an issue with your test run results, as I did here for Fedora 43 AArch64 a couple months ago.