testing#40
Closed
Proletter wants to merge 1 commit into
Closed
Conversation
ogad-tether
added a commit
that referenced
this pull request
Jun 8, 2026
Companion to the tts-cpp PR #40 memory-cycle fix. Constructs + destroys 20 `tts_cpp::supertonic::Engine` instances back-to-back, synthesising once per engine, and asserts that resident memory does not drift more than 5 MB across cycles 2..20 vs cycle 1. First cycle excluded because it captures one-time process-singleton inits (backend registry load, metal library compile, OpenMP pool, allocator arenas) that subsequent cycles reuse. Gated behind QVAC_TEST_SUPERTONIC_GGUF, same pattern as the existing SupertonicRealGguf round-trips in test_supertonic_config.cpp. Runs in `cpp-test-coverage-tts-ggml.yml` (which already provisions and exports the env var) so the on-PR workflow now exercises the fix end-to-end with an RSS assertion. Threshold (5 MB) absorbs allocator noise but catches a regression of a single leaked gallocr per cycle (each ~83 KB; 30 caches per synth -> ~2.5 MB per cycle, visible by cycle 3). Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
ogad-tether
added a commit
that referenced
this pull request
Jun 8, 2026
Rolls cpp-test-coverage-tts-ggml.yml back to upstream/main's single `ubuntu-22.04 / linux / x64` matrix entry. The macos-14-xlarge + ubuntu-24.04-arm additions in 209d299 both failed the cmake build on the GH-hosted runners: - linux-arm64: vcpkg unable to detect the active compiler (cc / c++ toolchain missing on the github-hosted ubuntu-24.04-arm image — needs an explicit `apt-get install -y build-essential` that the existing workflow flow doesn't run). - darwin-arm64: tts_ggml_tests compiles + links cleanly, but `bare-make build` runs an `install` step that expects `qvac__tts-ggml@0.bare` to exist (a Bare runtime artefact the addon build produces) and that artefact is missing because the coverage build target only built `tts_ggml_tests`. Both are fixable but neither is a small change — proper multi- platform support needs the `setup-vcpkg` reusable action (which the embed/llm workflows use), a bare-runtime install gate, and a Linux-arm64 toolchain provisioning step. Out of scope for the PR #40 memory-cycle fix's coverage gate; tracked as a follow-up. The linux-x64 entry continues to run `SupertonicRealGguf.NoLeak OverEngineCycles` against the downloaded supertonic.gguf, which is the leak-regression gate the PR needs. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.