Co-authored-by: Renovate Bot <[email protected]>

Co-authored-by: Renovate Bot <[email protected]>

Co-authored-by: Renovate Bot <[email protected]>

Co-authored-by: Renovate Bot <[email protected]>

Co-authored-by: Renovate Bot <[email protected]>

Co-authored-by: Renovate Bot <[email protected]>

Co-authored-by: Renovate Bot <[email protected]>

Co-authored-by: Renovate Bot <[email protected]>

… to v10.0.302 (#6407)

Co-authored-by: Renovate Bot <[email protected]>

…10 (#6410)

Co-authored-by: Renovate Bot <[email protected]>

Co-authored-by: Renovate Bot <[email protected]>

Co-authored-by: Renovate Bot <[email protected]>

…plateverifier to 10.0.302 (#6411)

Co-authored-by: Renovate Bot <[email protected]>

Co-authored-by: Renovate Bot <[email protected]>

Co-authored-by: Renovate Bot <[email protected]>

Co-authored-by: Renovate Bot <[email protected]>

Co-authored-by: Renovate Bot <[email protected]>

…HTML report + single GitHub summary (#6415)

* feat: cross-process test report aggregation (#4522)

Multiple test projects run as separate processes, producing N summary
blocks and N HTML reports with no way to merge them. Each process now
persists a JSON sidecar of its report data; sibling processes
cooperatively rewrite one marker-delimited summary block and one merged
HTML report under a file lock (last finisher leaves the complete
aggregate). On by default where a shared directory is resolvable
(GitHub Actions or TUNIT_AGGREGATE_DIR); TUNIT_AGGREGATE_REPORTS=off to
disable, =defer for multi-step pipelines with the new tunit-report
dotnet tool (TUnit.Reporting.Tool) merging at the end.

* docs: cross-job and cross-run report merging recipe

Full artifact-based workflow: per-job sidecar upload (defer mode),
download-artifact with merge-multiple onto one runner, tunit-report
merge for the combined HTML + single summary block; plus gh run
download for combining separate workflow runs.

* refactor: apply code-review cleanups to report aggregation

Single lock cycle: HtmlReporter now drives sidecar write, merged HTML
and the summary region in one env-parse/lock/scan pass, calling back
into GitHubReporter for the region write; previously both reporters
independently parsed env, locked and re-read every sidecar. Sidecar
JSON serializes once to UTF-8 bytes for both copies. Shared AtomicFile
helper replaces two swap implementations. ReportTestResult is a record
so merging uses 'with' instead of a drift-prone manual copy. Sidecar
reader collapses eight loops into a generic ReadArray. Tool links real
constants via new constants-only partials (TUnitActivitySource,
ClassTimelineAttribute) instead of value-mirroring shims; merged
report name and summary cap live in linked files. Duration formatting
and the failures-per-group cap are shared between the per-suite and
aggregated summaries.

* fix: harden sidecar reading against corrupt and unreadable files

Review findings (Codex + Claude): GetInt32() on a non-Int32-representable
schemaVersion threw FormatException past the catch(JsonException) guard,
letting one corrupt sidecar kill the whole merge; the tool's per-file
read was unguarded, so a locked or unreadable sidecar exited the tool
with code 2 instead of counting as skipped. schemaVersion and the int
helpers now use TryGetInt32; the tool skips unreadable files like the
engine does.

* fix: address review findings on merge fidelity and sidecar robustness

Sidecar reader now treats wrong-shaped nested values (groups:[1] throws
InvalidOperationException, not JsonException) and non-finite numeric
literals (1e9999 parses to Infinity) as malformed input instead of
letting them escape or poison duration math. Merged reports re-tag
tunit.test.class span tags when class names are disambiguated, so
per-class timelines survive multi-TFM merges. Source-control metadata
and link templates are kept only when every suite agrees — merging
sidecars from different commits no longer links tests to the wrong
revision. SpanData is a record to support the retag via 'with'.

* fix: tolerate inaccessible subdirectories during sidecar discovery

SearchOption.AllDirectories throws mid-enumeration when a subdirectory
is unreadable or deleted during the walk, exiting the tool before the
per-file guard runs. EnumerationOptions with IgnoreInaccessible skips
such directories instead.

* fix: correct suite-span rename key and dedupe engine sidecar reads

The class-timeline join resolves a suite span's class via
FindTagValue(TagTestClass) ?? Name, and suite spans carry no
TagTestClass — the collector rewrites their Name from test.suite.name.
The previous retag targeted TagTestClass (FullName values that never
match group names), so it was a no-op; renames now rewrite the suite
span's Name and test.suite.name tag, matching the report's actual join
key. Engine ReadAllSidecars also dedupes by content hash so pointing
TUNIT_AGGREGATE_DIR at a directory that already holds the local sidecar
(e.g. TestResults) no longer double-counts the suite.

* fix: pack reporting tool with build; replace stale block on summary overflow

CI failed with NETSDK1004: PackTUnitFilesModule packs --no-build, but
TUnit.Reporting.Tool is referenced by no test module so nothing had
restored or built it — pack now builds that one project (it isn't
strong-named, so the no-build race the module guards against doesn't
apply). Also, when a later aggregate exceeds GitHub's 1MB summary cap,
the previously written partial block is swapped for a short overflow
notice instead of being left standing and misreporting which suites
ran (review finding).