fix(coverage): handle all-zero calls in C-size reports #1316
No reviewers
Labels
No labels
area/accessibility
area/audio
area/input
area/install
area/menu
area/renderer
area/saves
area/tooling
area/upscaling
beanstalk: approved
beanstalk: failed
beanstalk: needs approval
beanstalk: running
beanstalk: tested
ci: approved
good first issue
gpu/amd
gpu/igpu
gpu/intel
gpu/nvidia
help wanted
kind
bug
kind
chore
kind
docs
kind
feature
kind
performance
kind
question
needs measurements
platform/linux
platform/macos
platform/steam-deck
platform/windows
preview
priority
critical
priority
high
priority
low
regression
risk
high
risk
low
risk
medium
status
blocked
status
confirmed
status
duplicate
status
invalid
status
needs-info
status
wontfix
ci: approved
No milestone
No project
No assignees
2 participants
Notifications
Due date
No due date set.
Dependencies
No dependencies set
Reference
MagicBeanDev/Longshot!1316
Loading…
Reference in a new issue
No description provided.
Delete branch "NitnekLeBorgne/Longshot-contributions:fix/coverage-zero-call-csize-20261117"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Type: fix · Risk: low · Tested on: Linux and native Windows, game-free synthetic captures; no GPU/driver exercised · Needs: none
Why
A valid coverage snapshot with known functions but zero calls crashes the offline report when C-size data is supplied. Such a snapshot is valid under the producer format; it is not an empty universe of functions.
The issue tracker was searched again on 2026-11-17 for
--csize,zero-callandzero call; no matching issue was found in the returned observational corpus. Hidden/deleted issues are not covered; no issue closure is claimed.How to reproduce
Seen on: Linux and native Windows, offline fixtures; GPU and driver not involved.
Settings and switches: Private synthetic fixtures; no game settings or GAME_PATHS inputs.
Steps: Run tests/test_coverage_report.py with its synthetic all-zero capture and C-size fixture, including the JSON variant in the independent replay. Run
python -I -B tests/test_coverage_report.pyagainst baseline production plus the added tests, then the combined candidate.Expected: Report generation succeeds with zero covered/cost values, empty windows and covered.txt, and no invented percentiles.
Actual: The baseline indexes an empty percentile distribution and exits 1 with IndexError.
What changed
Calculate C-size percentiles only when buckets is nonempty. Retain the nonzero calculation and strict >0.1 cost threshold. The patch is limited to the following production/test pair, both Git mode
100644.scripts/coverage/cover.py— C-size percentile block guarded by and buckets.tests/test_coverage_report.py— supported failure and neighboring controls.Rebased onto main
544229a260c5f5008f241b9cf48ae6b01c1602b9. Test-only commit:4d2b1d3dd99fe37b52aa04a3174d26c5f64678f8. Final fix head:5b6ecbb9e6f5c9b4828862968450fbcaacfde30f. Production/test bytes and modes remain identical to the independently reviewed source. Full new trees and parent chains were independently verified; current-base normal qualification uses this actual HEAD.Measurements
Nothing to measure: this is a contributor-report correctness fix, not a gameplay, rendering or performance change.
Flags and settings
None
Testing
Tested: Linux: The targeted script passes with test_coverage_report: OK (exit 0, empty stderr); this script has zero unittest methods. SPEC and QUALITY each verify 7 scenarios on both baseline/candidate plus script/mutation runs; five neighboring scenarios are byte-identical. Native Windows: 1 complete top-level script PASS;
test_coverage_report: OK, native exit 0. Not a unittest-method count. Public scripts ran withpython -I -Band owned fixtures. Native receipts: Job final 0, complete stdout/stderr, no controller errors; all 818 source paths unchanged. V5's three native controls and 60-child burst passed independent verification. Validated short-lived-process metadata gaps remain: no universal visibility or filesystem/network sandbox claim. The native proof was acquired at original headd35fbca8930186ceaf77093b5a63d9c48d80200d; that production/test dependency closure and modes are unchanged by this rebase. It is retained evidence, not a new Windows run. This is a top-level script with internal CLI assertions, not a new unittest method count.Fresh normal Linux qualification on the pinned current base and actual rebased candidate used
python -I -B tests/run_all.py --jobs 2 --junit ..., identical interpreter/environment policy and real Git HEAD/index:544229a260c5f5008f241b9cf48ae6b01c1602b9: 721 checks, 571 PASS / 42 FAIL / 108 SKIP, native exit 1.5b6ecbb9e6f5c9b4828862968450fbcaacfde30f: 721 checks, 571 PASS / 42 FAIL / 108 SKIP, native exit 1.tests/test_coverage_report.pynow also exercises the zero-call report. No new normal-runner check identity is claimed. All 721 common statuses, skip reasons and complete failure payloads agree after narrowly recorded diagnostic-noise handling: owned source/run paths, source-proven temporary file/directory suffixes, unittest elapsed times, one GCC intermediate object and one synthetic-fixture backup directory. Assertion values, failure identities, symbols, numeric offsets and statuses are not normalized. The GCC/linker error still nameschoose_face,hos_options_file,.text+0xa1aand exit 1; a separate retained compiler/assembler/linker trace proves this toolchain's temporary-object provenance. No universal no-regressions claim.The 42 normal failures also observed on this current-main qualification are named here, as the template permits; the normal suite is not green:
tests/test_bug_report.pytests/test_bug_report_defaults.pytests/test_cpu_report.pytests/test_crash_groups.pytests/test_crt_options.pytests/test_damage_debug_menu.pytests/test_discover.pytests/test_display_startup_failure.pytests/test_dlss_feature_size.pytests/test_failed_compile_keeps_objects.pytests/test_framegen_preset.pytests/test_guest_imported.pytests/test_hardware_report.pytests/test_hud_layout_row.pytests/test_invalidate_vcore.pytests/test_lift_lookups.pytests/test_link_skip.pytests/test_linux_pressure.pytests/test_linux_update.pytests/test_longshot_saves.pytests/test_mirror_budget.pytests/test_missing_nup.pytests/test_mods_size_table.pytests/test_pending_guest_import.pytests/test_pipeline_play_cache.pytests/test_present_thread.pytests/test_profiling_mode.pytests/test_readback_timeout_source.pytests/test_renderer_new_handler.pytests/test_rumble_menu.pytests/test_run_all_changed.pytests/test_slow_frames.pytests/test_switch_table.pytests/test_target_image_pool.pytests/test_tested_configurations.pytests/test_upscaler_round_trip.pytests/test_upscaler_rows.pytests/test_vcore_escape.pytests/test_vk_occlusion_queries.pytests/test_vk_swapchain_compat.pytests/test_walk.pytests/test_walk_key.pyNot tested: Full Windows suite, real SDK/release installation, gameplay/GPU or hosted-CI equivalence. These source/fixture-only limits do not invent extra native-test obligations. Hosted checks remain tracked separately; readiness is not approval, a full CI PASS or merge eligibility.
Must be tested: None additional for this two-file contributor-tool change. The normal-runner exception is explicitly limited to the named current-main failures above, with source-only context disclosed; it is not a waiver for a new candidate failure or an unknown comparison.
Risks
What could break: Empty-distribution handling only. Empty function universes are outside the minimal contract. Evidence is fixture-only: no capture, instrumentation, injection or actual game session.
Turning it off: Revert the fix and its associated test commit.
Compatibility: No game runtime, saves, options, shader/pipeline cache or vk ABI changes; only the stated contributor-tool behavior changes. No GAME_PATHS directory is touched and no in-game run is claimed.
Checklist
The normal runner remains native exit 1; its named current-main failures use the template's explicit exception, not a full-suite PASS claim. Tracker search and the GPL-3.0-or-later diff/provenance check are complete. No incompatible dependency, third-party code, dump or decompiled/generated game content is introduced.
Everyone
type(scope): description) of the status line's type.can't be run by hand is in Testing's Not testable by hand, with why.
in Risks.
main.python -I -B tests/run_all.pypasses (any failure that also fails onmainis named).third_party/untouched (a new folder with aPROVENANCE.mdinstead).When it applies
Assisted-by:/Co-Authored-By:trailers.mainand passing here.tests/that reads changed source updated in this PR, not weakened.The retained RED uses identical baseline production and regression-test bytes; the rebased normal run verifies the final source on the current-base context. No host C/C++, shader, game switch or visible-game change applies.
d35fbca8935b6ecbb9e6Rebased onto main
544229a260c5f5008f241b9cf48ae6b01c1602b9with both reviewed file bytes and Git modes unchanged. New head:5b6ecbb9e6f5c9b4828862968450fbcaacfde30f. Fresh source-only Linux normal qualification matches current main at 571 PASS / 42 named FAIL / 108 SKIP, both native exit 1; all 721 common statuses and full failure/skip diagnostics agree after narrowly documented diagnostic-noise handling. The existing coverage check includes the new zero-call assertions; no extra normal-runner test ID is claimed. The prior native Windows top-level script proof remains tied to the unchanged source closure, not a new game capture or full-CI result.WIP: fix(coverage): handle all-zero calls in C-size reportsto fix(coverage): handle all-zero calls in C-size reportsBrought up to the ready-for-review standard (#1454): the checklist's Everyone list has two new boxes, added here from what this PR's description and comments show (please check they match what you know).
View command line instructions
Checkout
From your project repository, check out a new branch and test the changes.