Fix integer truncation in cosmic livetime resampling scale factor

[oksuzian]

Problem

Mu2eProductMixer::endSubRun computes the cosmic livetime scale factor with integer division:

float scaling = resampledEvents_ / generatedEvents_;

resampledEvents_ and generatedEvents_ are both unsigned int, so the ratio is truncated to its floor before the float assignment. The resampled CosmicLivetime product (instance mixed, the only livetime product S2Resampler.fcl keeps) is scaled by that floored value.

Consequences, measured against sim.mu2e.CosmicDSStopsCRYAll.MDC2025ab (GenEventCount = 128,783/subrun, S1 livetime 29.586 s/subrun):

dataset	events/job	true scaling	stored	livetime/subrun stored (true)
dts.mu2e.CosmicCRYAll.Run1Ban	100,000	0.776	0	0 s (22.97 s)
dts.mu2e.CosmicCRYAll.MDC2025ap	500,000	3.882	3	88.76 s (114.9 s, −23%)

Any S2 cosmic resampling with events/job below the mixin subrun's generated count stores exactly zero livetime; above it, livetime is understated by up to 1/floor(ratio). Present since c30db7e (2021).

Fix

Cast before dividing:

float scaling = static_cast<float>(resampledEvents_) / static_cast<float>(generatedEvents_);

One line; no interface change. Existing datasets can be corrected offline since the true scaling is recomputable from events/job and the S1 subrun GenEventCount.

🤖 Generated with Claude Code

[FNALbuild]

Hi @oksuzian,
You have proposed changes to files in these packages:

• EventMixing

which require these tests: build.

@Mu2e/write, @Mu2e/fnalbuild-users have access to CI actions on main.

⌛ The following tests have been triggered for 52ff4a0: build (Build queue - API unavailable)

About FNALbuild. Code review on Mu2e/Offline.

[FNALbuild]

☀️ The build tests passed at 52ff4a0.

Test	Result	Details
test with	✅	Command did not list any other PRs to include
merge	✅	Merged 52ff4a0 at 8ca6473
build (prof)	✅	Log file. Build time: 04 min 21 sec
ceSimReco	✅	Log file.
g4test_03MT	✅	Log file.
transportOnly	✅	Log file.
POT	✅	Log file.
g4study	✅	Log file.
cosmicSimReco	✅	Log file.
cosmicOffSpill	✅	Log file.
ceSteps	✅	Log file.
ceDigi	✅	Log file.
muDauSteps	✅	Log file.
ceMix	✅	Log file.
rootOverlaps	✅	Log file.
g4surfaceCheck	✅	Log file.
trigger	✅	Log file.
check_cmake	✅	Log file.
FIXME, TODO	✅	TODO (0) FIXME (0) in 1 files
clang-tidy	➡️	2 errors 14 warnings
whitespace check	✅	no whitespace errors found

N.B. These results were obtained from a build of this Pull Request at 52ff4a0 after being merged into the base branch at 8ca6473.

For more information, please check the job page here.
Build artifacts are deleted after 5 days. If this is not desired, select Keep this build forever on the job page.

[rlcee]

did you check for other instances of this issue in this module?

[oksuzian]

Yes — I swept the module. This is the only integer division in EventMixing: the only other / operators are in ProtonBunchIntensityLogNormal_module.cc and are all double-precision math.

I also checked the adjacent line: in totalPrimaries_ * scaling the product is a float, but the CosmicLivetime constructor's first argument is unsigned long long, so the primaries field is truncated to an integer. For CRY that field is a dummy (the generator stores primaries=1 and the CRY-internal livetime), and for CORSIKA the loss is under one primary out of the full count, so no change is needed there — the livetime itself flows through as float.

For scale of the original problem: with resampledEvents < generatedEvents the stored livetime is exactly 0 (e.g. dts.mu2e.CosmicCRYAll.Run1Ban, 100k resampled from ~129k available per subrun), and otherwise it is floored (dts.mu2e.CosmicCRYAll.MDC2025ap: scaling 3.88 stored as 3, livetime -23%; MDC2020ar: 11.82 stored as 11, -6.9%). Existing datasets can be corrected offline since the true ratio is recomputable from the dts and stage-1 gencounts.