Skip to content

client-suite.sh runs the tests of the client 2.18.1, and linux-tests.sh builds in memory - #16

Merged
Laurianti merged 2 commits into
mainfrom
local-scripts
Oct 29, 2026
Merged

Laurianti merged 2 commits into
mainfrom
local-scripts

Conversation

@Laurianti

Copy link
Copy Markdown
Member
  • build/client-suite.sh runs the tests of the last release of ChromaDotNet.Client, 2.18.1, instead of 2.14.0, from a clone for each tag in artifacts/client-: with a single artifacts/client, another CLIENT_TAG ran the clone of the tag before.
  • build/linux-tests.sh builds the tests in a copy of the repository in a tmpfs, not in the container in the disk of Docker, and of artifacts the copy takes only the packages and the clones of the client; CLIENT_TAG goes into the container. The .NET SDK and the packages of NuGet stay in their volumes: in folders of Windows mounted in the container, the build of the clone of the client and the first start of Chroma took 26 s and 403 ms, against 8 s and 104 ms.

Tests run:

  • build/client-suite.sh on Windows: against Chroma 1.5.9, 2176 passed, 12 skipped, 0 failed; against VecturaDB 1.0.0-preview.3, 2173 passed, 12 skipped, 0 failed. The script leaves out the three tests that expect a server without the Search API, the sparse vectors and the fork of collections: VecturaDB has them, so these tests fail against it.
  • build/linux-tests.sh ubuntu:24.04 ubuntu2404 x64: the tests of ChromaDotNet.Local, 90 passed, 4 skipped (the two tests for Windows only, for each engine), 0 failed; the suite of the client against Chroma 1.5.9, 2176 passed, 12 skipped, 0 failed.

…8.1, from a clone for each tag

The suite ran the tests of ChromaDotNet.Client 2.14.0; it now runs those of 2.18.1, the last release. On Windows 2176 pass against
ChromaDB and 2173 against VecturaDB, which does not run the three tests that expect a server without the Search API, the sparse
vectors and the fork: they did not change in 2.18.1.

The clone goes into artifacts/client-<tag>. With a single artifacts/client, the clone of the tag before stayed, and another
CLIENT_TAG ran the same tests. The log says the tag of the tests.
…, not in the disk of Docker

The copy of the repository where the tests build was in the container, in the disk of Docker, and took all of artifacts but the
cache of NuGet, with the outputs of the other scripts. It is now in a tmpfs of 4 GB, and of artifacts it takes only the packages
and the clones of the client.

The .NET SDK and the packages of NuGet stay in their volumes. In folders of D: mounted in the container, which reads them through
the file sharing of Docker, the build of the clone of the client, as the seconds of the suite less the duration of its tests, and
the first start of Chroma were slower. On ubuntu:24.04 x64, with the copy in memory and little load on the machine, they took:
- SDK and packages in folders of D:: 26 s and 403 ms
- SDK and packages in memory too: 7 s and 136 ms
- SDK and packages in the volumes: 8 s and 104 ms

CLIENT_TAG goes into the container, for build/client-suite.sh to run the clone of the same tag.
@Laurianti
Laurianti requested a review from a team as a code owner October 29, 2026 18:12
@Laurianti
Laurianti merged commit bb19bcc into main Oct 29, 2026
13 checks passed
@Laurianti
Laurianti deleted the local-scripts branch November 7, 2026 10:19
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants