test_search_result_context_calculation fails on x86 darwin #3

New Issue

2026-02-16T04:15:05-05:00

yindo commented

2026-02-16 04:15:05 -05:00

Originally created by @amadejkastelic on GitHub (Sep 3, 2025).

Ran across the following failure on x86 darwin when trying to package semtools in nixpkgs:

test tests::test_search_documents_with_max_distance ... ok
test tests::test_search_result_context_calculation ... FAILED

failures:

---- tests::test_search_result_context_calculation stdout ----

thread 'tests::test_search_result_context_calculation' panicked at src/bin/search.rs:322:13:

assertion left == right failed

left: 2

right: 3

note: run with RUST_BACKTRACE=1 environment variable to display a backtrace

failures:

tests::test_search_result_context_calculation

test result: FAILED. 7 passed; 1 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.01s

error: test failed, to rerun pass --bin search

Originally created by @amadejkastelic on GitHub (Sep 3, 2025). Ran across the following failure on x86 darwin when trying to package semtools in nixpkgs: ``` test tests::test_search_documents_with_max_distance ... ok test tests::test_search_result_context_calculation ... FAILED failures: ---- tests::test_search_result_context_calculation stdout ---- thread 'tests::test_search_result_context_calculation' panicked at src/bin/search.rs:322:13: assertion left == right failed left: 2 right: 3 note: run with RUST_BACKTRACE=1 environment variable to display a backtrace failures: tests::test_search_result_context_calculation test result: FAILED. 7 passed; 1 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.01s error: test failed, to rerun pass --bin search ```

yindo closed this issue

2026-02-16 04:15:05 -05:00

yindo commented

2026-02-16 04:15:05 -05:00

@logan-markewich commented on GitHub (Sep 3, 2025):

Hmm, I think in some machines the similarity function can be slightly different. Since im using constant vectors, floating point error has measurable impact.

I'll redesign the test to account for this

@logan-markewich commented on GitHub (Sep 3, 2025): Hmm, I think in some machines the similarity function can be slightly different. Since im using constant vectors, floating point error has measurable impact. I'll redesign the test to account for this

yindo commented

2026-02-16 04:15:05 -05:00

@amadejkastelic commented on GitHub (Sep 3, 2025):

Here is the run: https://github.com/amadejkastelic/nixpkgs-review-gha/actions/runs/17431732970/job/49491508641
It was run without modifying any code. I'll also look into the code.

@amadejkastelic commented on GitHub (Sep 3, 2025): Here is the run: https://github.com/amadejkastelic/nixpkgs-review-gha/actions/runs/17431732970/job/49491508641 It was run without modifying any code. I'll also look into the code.

yindo commented

2026-02-16 04:15:06 -05:00

@logan-markewich commented on GitHub (Sep 3, 2025):

@amadejkastelic I think the above PR should fix it -- decided that the embedding model is lightweight enough to just use it in tests. Should be more realistic imo

@logan-markewich commented on GitHub (Sep 3, 2025): @amadejkastelic I think the above PR should fix it -- decided that the embedding model is lightweight enough to just use it in tests. Should be more realistic imo

yindo referenced this issue

2026-02-16 04:15:14 -05:00

[PR #3] [MERGED] unify packaging into one crate-multi-binary #18

Sign in to join this conversation.

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: run-llama/semtools#3