Repository navigation
p110 for kinkal 3.9.1 - #2100
Merged
Merged
Conversation
Collaborator
|
☀️ The build tests passed at 7bccd29.
N.B. These results were obtained from a build of this Pull Request at 7bccd29 after being merged into the base branch at 5629b9f. For more information, please check the job page here. |
oksuzian
approved these changes
Nov 4, 2026
oksuzian
left a comment
Collaborator
There was a problem hiding this comment.
PR Review Summary — #2100
Reviewed at head 7bccd295526d1ce40ca60773e5e05a91ad7b0b24. First pass.
Decision
- 🟢 approve. Moving from p109 to p110 changes only KinKal, from v3.9.0 to v3.9.1; the other changed build, cry 1.7, is rebuilt from identical sources. FNALbuild built this head with p110, and every test passed. On the nightly
reco_03sample, the change in reconstruction output is exactly the one I measured for KFTrack/KinKal#231 before it merged. Offline#2084 is waiting for this envset, so this PR should merge first.
Scope understood
- One line:
.musenow saysENVSET p110instead ofp109. - The p109 and p110 envset files differ in two lines: the spack environment name, and
KINKAL_VERSION(v03_09_00becomesv03_09_01). The two spack environments list the same packages. Only two builds differ: KinKal, and cry 1.7 (see below). Art, ROOT, Geant4, gcc and artdaq-core-mu2e are the same builds. - KinKal v3.9.1 is v3.9.0 plus two merged PRs, in three headers (+41 −2):
- KFTrack/KinKal#231: a fallback in
PiecewiseTrajectory::t0()for when the search finds no self-consistent piece. - KFTrack/KinKal#232:
Track::coverEndDomain, which extends the end piece to the outermost field domain before a field-corrected extrapolation.
- KFTrack/KinKal#231: a fallback in
Findings
- 🟡 [S2] The description does not say that p110 changes reconstruction output. On
reco_03, most downstream fits save more segments, and several hundred report a differentt0and momentum (numbers below). Anyone who sees these shifts in validation after the merge will look for a cause. Please add one sentence, for example: "KinKal 3.9.1 changes which trajectory piece sets t0; most downstream KKDe/KKDmu fits save more segments, and a few report t0 up to 1.8 ns and momentum up to 2.4 MeV/c different."
Checked, no action needed
- 🟢 FNALbuild build 3551 is a real p110 build. All 1,714
g++lines in itsscons.logwrite tobuild/al9-prof-e29-p110, and the log never mentions p109. Every row of the result table passed, including the eleven validation jobs,trigger,check_cmake, both overlap checks and clang-tidy. - 🟢 Reconstruction output, p109 against p110, on all 16,508 events of
Production/Validation/nightly/reco_03.fclwith its two dig inputs.- Setup. The p109 run used the CI build of
main5629b9f2, which is this PR's merge base. For p110, I built this head withmuse build, with the same Production (94a4d81f) and mu2e-trig-config (5999ffa2). The build finished with exit status 0 and no warnings. EveryKalSeedcollection was printed for every event. All 12 jobs exited with status 0, with no exceptions. - The same in both runs: the track count in every collection, every fit flag, fit consistency, hit and straw counts, and every crossing list except one track. All trigger collections,
KKUmu, and everyKKUetrack but one are identical. KKDe: 13,290 of 16,491 tracks save more segments (median 13 more, never fewer). 68 report a different momentum (at most 0.15 MeV/c), and 619 a differentt0(at most 0.4 ns).KKDmu: 13,179 of 16,316 tracks save more segments (median 16 more, never fewer). 100 report a different momentum (2 by more than 1 MeV/c, at most 2.4 MeV/c), and 778 a differentt0(2 by more than 1 ns, at most 1.8 ns).KKUe: one track, in event 1430:1652:852, saves 106 more segments and gains 4 calorimeter crossings. Itst0of 531.7 ns now lies inside its range. This is the event behind Offline#2076.- Both runs are byte-identical, chunk by chunk, to the runs from my KinKal#231 review: v3.9.0, and #231 at
69375ef3, both on Offline68b28d6e. So two things after that head change nothing on this sample: #231's last commit (iteration limit back to 10, simpler convergence test) and #232. Nor do themainchanges since68b28d6e. That review traced the cause. After a downstream fit is extrapolated upstream, its front piece is usually the reflected leg, whoset0lies before the trajectory. The v3.9.0 search stops on that piece, and v3.9.1 converges on the tracker piece.KKFit::createSeeduses the direction att0()to choose which tracker-boundary crossings bound the saved segments. So it now keeps the segments betweenTT_Frontand the first hit.
- Setup. The p109 run used the CI build of
- 🟢 cry. p110 uses a different build of cry 1.7 (
vlstp2b) from p109 (z6psfn5). It is the build that p107 and p108 used.diff -rqof the two install trees finds identical sources, headers and data files. The spack recipe differs only in its import line, and the library was rebuilt from the same sources. - 🟢 Nothing else needs a matching bump. No other in-scope repo pins an envset in a
.musefile: Production, mu2e-trig-config, EventNtuple, PassN, EventDisplay, DQM and RefAna. The CMake path finds KinKal without a version (CMakeLists.txt:74). - 🟢 Merge order. Offline#2084 asks to merge "once a KinKal tag with KFTrack/KinKal#232 is in the envset" (its Dependency section). This PR provides that tag. No other open Offline PR touches
.muse.
Validation check
- Build/tests run: yes. I ran one full
muse buildof this head with p110, and the 16,508-event comparison above. I also ran a standalonet0()test against the p110 KinKal headers and libraries. - Config contract check: not applicable. The PR changes no fcl and no code.
- Cross-repo consistency: pass.
Residual risk
- In the released
t0(), the closest-piece fallback runs only when the search ends outside the trajectory. Suppose the search cycles between two pieces and no piece is self-consistent. Then it returns what v3.9.0 returned.- In my standalone test, pieces with
t0values 1.5, 2.9, 1.1 and 4.3 on [0,4] give 1.1 with both v3.9.0 and v3.9.1. KinKal#231 at69375ef3gave 4.3. - On
reco_03the output matches69375ef3exactly, so this case either does not occur there or does not change the result. This is a question for KinKal, not for this PR.
- In my standalone test, pieces with
- Not checked:
- Off-spill cosmics and
CentralHelixfits with p110, beyond FNALbuild'scosmicOffSpillandcosmicSimRecojobs. My KinKal#232 review found the productionKKCHmusetting (field correction off) unchanged by #232, but I did not rerun it with p110. - A CMake build of Offline with p110.
- Off-spill cosmics and
Author follow-ups
- Add one sentence to the description about the change in reconstruction output (finding 1).
brownd1978
approved these changes
Nov 4, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Upate muse envset for new kinkal version, 3.9.1.
This changes the t0 value for some track fits, and saves more segments.