• Home
  • Features
  • Pricing
  • Docs
  • Announcements
  • Sign In

kubernetes-sigs / inference-perf / 42967643708
87%

Build:
DEFAULT BRANCH: main
Ran 11 Nov 2026 05:55AM UTC
Jobs 1
Files 141
Run time 1min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst

11 Nov 2026 05:43AM UTC coverage: 86.286% (+0.003%) from 86.283%
42967643708

push

github

web-flow
Address the review items left on #919 (#935)

Part of #449

Follow-up to the review on #919. Every item there reproduced on main
except half of one: a `_term0 > _term0` comparison does not fold, so
only the `0 * term` case needed the pruning fix. Nothing on main
constructs a `Predicate` yet, so no runtime behaviour changes for a
user.

Two commits. The first addresses the review items. The second drops the
solver cross-check, which is where two of the items lived.

Changes:

* A `t > 0` gate was rejected: the earliest-boundary cross-check probed
`0 * 1.001`, which is 0, outside the open interval.
* The cross-check is gone (second commit). `(t >= 600) | ((t > 0) &
(errors > 10))` proves to (600, 0), pinned by a test.
* An early-abort window like `(t < 300) & (errors > 10)` is rejected,
and the rejected-condition table did not say so.
* Added the row. The rejection stays: a result-driven check can miss a
window that does not start at 0, so unions are not accepted.
* The boundary cross-check no longer probed the compiled `_holds`, so
NaN-makes-false was assumed rather than checked.
* A test pins `holds(boundary, {})` and `holds(0.999 * boundary, {})` on
a term condition. No code change: the check that would have carried it
is gone.
* A term sympy folded away stayed in `terms`, so a `MeasuredState` would
evaluate a window the condition never reads.
* `_terms` is pruned to the placeholders left in the parsed expression.
`terms` and `measured_symbols` now agree.
* A parse error named the rewritten string, so `_term0` leaked into the
message instead of the author's spelling.
  * `_parse_raw` takes `shown`, the source text to name in the error.
* count, min, max and quantile windows maintained the exact-integer sums
they never read, about 1.1 us per sample.
* `_Window` keeps the sums only for sum, avg and stddev, the way it
already keeps the sorted list only for min, max and quantile.
* The docs said terms read "the same per-request values the report is
compu... (continued)

18 of 18 new or added lines in 2 files covered. (100.0%)

1 existing line in 1 file now uncovered.

13716 of 15896 relevant lines covered (86.29%)

0.86 hits per line

Coverage Regressions

Lines Coverage ∆ File
1
79.17
-1.04% inference_perf/datagen/base.py
Jobs
ID Job ID Ran Files Coverage
1 42967643708.1 11 Nov 2026 05:55AM UTC 141
86.29
GitHub Action Run
Source Files on build 42967643708
  • Tree
  • List 141
  • Changed 3
  • Source Changed 2
  • Coverage Changed 3
Coverage ∆ File Lines Relevant Covered Missed Hits/Line
  • Back to Repo
  • Github Actions Build #42967643708
  • 979e69eb on github
  • Prev Build on main (#42925197733)
  • Next Build on main (#42967863938)
  • Delete
STATUS · Troubleshooting · Open an Issue · Sales · Support · CAREERS · ENTERPRISE · START FREE TRIAL · SCHEDULE DEMO
ANNOUNCEMENTS · TWITTER · TOS & SLA · Supported CI Services · What's a CI service? · Automated Testing

© 2026 Coveralls, Inc