Skip to content

buildtest: remove --long-running/--parallel-algorithm flags that break benchmark test discovery #723

Description

@credfeto

Prompt (verbatim, from live chat)

can you also apply this change to the buildtest in scripts?

Context: while fixing credfeto/credfeto-global-pre-commit#189 (per-project benchmark dotnet test invocation passing --long-running 10 --parallel-algorithm sequential, which a real benchmark project's test host rejects as invalid arguments, exit code 5, "Zero tests ran"), the human asked for the equivalent flag removal in this repo's development/buildtest, which has its own (older, single-phase) dotnet test invocation with the same two flags.

Relationship to #679

#679 is already open in this repo for a related but broader problem: --parallel-algorithm aggressive racing benchmark assemblies against regular unit tests causes OutOfMemoryException. Its pending Implementation Plan comment (from a different agent identity) proposes a larger refactor: making benchmark exclusion the default, fixing the --no-benchmmarks/--no-integratopn flag typos, and fixing the broken TEST_BENCHMARKS=--filter-not-namespace ... unquoted assignment.

This issue is intentionally narrower: it only removes the two flags shown to cause outright argument-rejection failures (confirmed by direct reproduction against FunFair.Ethereum.DataTypes.Benchmark.Tests), matching the fix already applied in credfeto-global-pre-commit#189. It does not attempt the broader refactor or typo fixes tracked in #679's pending plan, to avoid conflicting with that in-flight work.

Proposed fix

In development/buildtest, remove --long-running 10 and --parallel-algorithm aggressive from the single dotnet test invocation (current lines 169 and 172).

Evidence

Direct reproduction in funfair-server-ethereum: running the equivalent dotnet test invocation without --long-running 10 --parallel-algorithm sequential (the analogous flags in the two-phase credfeto-global-pre-commit version) passed 4/4 tests in 4m20s; with them, the same project produced exit code 5 ("Zero tests ran") in under a second.

Metadata

Metadata

Assignees

Labels

AI-WorkWork for an AI Agent

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions