Last updated on 2026-10-08 00:54:23 CEST.
| Flavor | Version | Tinstall | Tcheck | Ttotal | Status | Flags |
|---|---|---|---|---|---|---|
| r-devel-linux-x86_64-debian-gcc | 0.8.1 | 3.96 | 74.70 | 78.66 | OK | |
| r-devel-linux-x86_64-fedora-gcc | 0.8.1 | 73.17 | OK | |||
| r-release-macos-arm64 | 0.8.1 | 1.00 | 30.00 | 31.00 | ERROR | |
| r-release-macos-x86_64 | 0.8.1 | 4.00 | 155.00 | 159.00 | OK | |
| r-oldrel-macos-arm64 | 0.8.1 | ERROR | ||||
| r-oldrel-macos-x86_64 | 0.8.1 | 4.00 | 130.00 | 134.00 | OK | |
| r-oldrel-windows-x86_64 | 0.8.1 | 12.00 | 152.00 | 164.00 | OK |
Version: 0.8.1
Check: tests
Result: ERROR
Running ‘core.R’ [16s/16s]
Running the tests in ‘tests/core.R’ failed.
Complete output:
> library(lambdastar)
>
> # Deliberately base R: the tests need no optional testing installation.
> .count <- 0L
> .failures <- character()
> test <- function(name, code) {
+ tryCatch({ force(code); .count <<- .count + 1L; cat("PASS:", name, "\n") },
+ error = function(e) {
+ .failures <<- c(.failures, paste(name, conditionMessage(e), sep = ": "))
+ cat("FAIL:", tail(.failures, 1), "\n")
+ })
+ }
> equal <- function(actual, expected, tol = 1e-10) {
+ result <- all.equal(actual, expected, tolerance = tol, check.attributes = TRUE)
+ if (!isTRUE(result)) stop(paste(result, collapse = "; "))
+ }
> relative <- function(actual, expected, tol = 1e-8) {
+ stopifnot(length(actual) == length(expected), all(is.finite(actual)),
+ all(is.finite(expected)), all(expected != 0),
+ all(abs(actual / expected - 1) <= tol))
+ }
> fails <- function(code, pattern) {
+ err <- tryCatch({ force(code); NULL }, error = identity)
+ if (is.null(err) || !grepl(pattern, conditionMessage(err)))
+ stop("Expected error matching: ", pattern)
+ }
> test_root <- if (dir.exists("cases")) "." else "tests"
> fixture <- function(name) file.path(test_root, "fixtures", name)
> for (file in list.files(file.path(test_root, "cases"), full.names = TRUE)) source(file)
PASS: measurement API preserves scale and case provenance
PASS: matrix, formula, lm and H1 paths agree
PASS: two-group adapter agrees with pooled-variance t-test
PASS: ANOVA, ANCOVA and interaction use the full encoded model
PASS: recoding preserves projection but can change geometry and F_H
PASS: lm stored contrasts survive changes to global contrast options
PASS: response mismatches and external formula variables are rejected
PASS: lm validates saved cases, design and response against common data
PASS: weights, offsets and other model families are rejected
PASS: rank deficiency, lost levels and invalid explicit row identity fail
PASS: mean response and F_H temperature stay coupled under item scaling
PASS: null model and temperature singularity remain explicit
PASS: required data, indicators, resolution and settings are validated
PASS: S3 summaries and printing expose estimates without full matrices
PASS: bootstrap defaults and explicit overrides are resolved and validated
PASS: same seed gives identical joint draws and changing seed changes draws
PASS: bootstrap preserves RNG kind and present or absent caller seed
PASS: joint resampling matches independent pairwise moments and lm.fit
PASS: percentile intervals use requested level and joint kappa ratios
PASS: missingness and strata use one original selection and retain row mapping
PASS: fixed-mixture variance requires estimable within-stratum variances
PASS: comparison shares draws with separate matrix/formula/lm pipelines
PASS: data-dependent transformations rebuild while factor coding stays fixed
PASS: lost factor levels fail only the affected model and never get replaced
PASS: zero temperature skips every CI including temperature without sampling
PASS: zero total structure never yields an invented kappa interval
PASS: item scaling transforms both point and interval noncentralities
PASS: retention is optional, thresholds explicit, summaries include diagnostics
PASS: numerical eta agrees with independent 90-digit Decimal reference
PASS: tiny eta follows analytic three-bin limit instead of 0.5
PASS: capacity scaling, vectorization and high-resolution behavior
PASS: thermal boundaries and smooth small-temperature stability
PASS: direct quadrature agrees with known primitive and small-ratio limit
PASS: lookup agrees with direct integration across range and off-grid
PASS: numerical controls are validated and limits fail explicitly
PASS: comparison equals separate evaluations and obeys swap/self identities
PASS: common missingness selection precedes both data-dependent designs
PASS: matrix missingness joins the same common complete-case selection
PASS: lm case omissions are accepted only when they match the common rows
PASS: two explicit models and a common matrix are required
PASS: comparison retains each formula coding and common numerical settings
PASS: transform-created missing and nonfinite values never alter cases silently
PASS: E3 preserves 0.5.0 whole and conditional adaptive-basis outputs
PASS: general multirow L agrees with independently fitted restricted models
PASS: coefficient, matrix and lm sources retain the same conditional geometry
PASS: full coefficient restrictions reduce to a whole-model state
PASS: L units and invertible recoding preserve capture but retain F_H geometry
PASS: L input errors and nonidentifiability are explicit
PASS: derived null and target are rebuilt for every complete-case bootstrap draw
PASS: contrast cluster draws preserve all rows and variable sample sizes
PASS: fixed and reevaluated polynomial/spline blocks reproduce their declared draw estimators
PASS: fixed adaptive nuisance and whole-model encoding are learned after selection
PASS: simple and explicitly averaged grid effects agree with named coefficient L
PASS: grid prediction uses learned poly, spline and scale coordinates
PASS: fixed grid contrasts use original learned basis with draw-specific restricted fits
PASS: fixed named polynomial coefficients differ from reevaluated coordinates as declared
PASS: grid validation prevents implicit weighting and unsupported prediction recipes
PASS: Fourier and explicitly fixed-knot piecewise/spline recipes equal encoded matrices
PASS: rank failures remain original draw failures without replacement
PASS: multirow grid recoding shares one null reference and retains factor coding
PASS: contrast singular temperature and undefined kappa retain their meanings
PASS: parallel measurements agree with independent pairwise identities
PASS: two-replicate identity and default composite scale
PASS: partition identity, scaling, and preserved observation identity
PASS: zero bath is singular while zero signal has undefined kappa
PASS: invalid measurements and row mismatches are rejected
PASS: tiny positive temperature is not clipped to zero
PASS: 0.4.1 baseline points, geometry, draws and intervals remain compatible
PASS: saved designs retain coding independently of global contrasts and caller functions
PASS: lm and matrix design objects preserve their encoded models
PASS: bound designs reject stale predictors and row order but permit new measurements
PASS: conditional blocks match restricted fits and actual encoded geometry on unequal cells
PASS: block transformations are rebuilt after shared selection and inside each draw
PASS: additional nuisance and conditional cluster bootstrap keep common row provenance
PASS: comparison requires explicit compatible references and preserves paired draws
PASS: conditional comparisons share nuisance while retaining distinct target columns
Tracing function ".nuisance_space" in package "namespace:lambdastar"
Tracing function ".thermal_terms" in package "namespace:lambdastar"
Untracing function ".nuisance_space" in package
"namespace:lambdastar"
Untracing function ".thermal_terms" in package "namespace:lambdastar"
PASS: shared point evaluation prepares nuisance and thermal terms only once
PASS: term selections, omissions and rank failures have explicit meaning
PASS: designs preserve singular temperatures and raw coding sensitivity
PASS: panel time trend matches explicit person dummy least squares
PASS: default panel occasion effect is categorical with literal coding
PASS: additional nuisance is projected within persons with correct rank
PASS: panel target blocks express group by occasion effects after person adjustment
PASS: person slopes rebuild separate copies in every bootstrap draw
PASS: unbalanced panel bootstrap retains each selected person's entire pattern
PASS: person resampling keeps stratum counts and rejects time-varying strata
PASS: panel complete selection removes whole persons on common comparison variables
PASS: panel comparisons reuse the same measurement and standalone numerical F
PASS: person fixed effects can be targets with distinct bootstrap identities
PASS: paired two-occasion reduction agrees with E6a lambda values and scaled F
PASS: panel scale and person-constant shifts preserve appropriate quantities
PASS: panel validation rejects ambiguous maps and incompatible estimator paths
PASS: panel fixed non-person bases and explicit matrices remain usable
PASS: declared person factor contrasts survive panel preparation and resampling
PASS: panel zero temperature and zero signal retain explicit diagnostic meanings
PASS: saturated person by occasion interactions retain positive bath df
PASS: unavailable library percentiles return diagnostics without substituted bounds
PASS: single-model F_H agrees with independent lm and determinant
PASS: recoding changes F_H solely through geometry
PASS: singular inference is distinct from the mathematical boundary
PASS: null model, smooth opt-in, and response-scale covariance
PASS: population estimator is independent of bootstrap and confidence level
PASS: every stored draw matches independent pairwise moments and OLS
PASS: percentile construction agrees with boot on the identical person draws
PASS: weak signal records undefined ratios without substituting an interval
PASS: valid zero/one ratios and degenerate distributions retain their percentiles
PASS: fixed-stratum population draws match independent mixture variance
PASS: null models have valid zero localization for positive signal
PASS: saturated models keep finite-person F_H and identify unestimable population projection
PASS: singular temperature draws preserve temperature zeros and report dependent failures
PASS: zero total point retains a lambda interval and an undefined kappa point
PASS: unrepresentable tail probabilities are reported without substituting limits
PASS: known orthogonal energy partition is 9 plus 16
PASS: encoded geometry differs while recoded projection agrees
PASS: null, one direction, rank loss, and bad projection
PASS: reduced factor contrasts agree across formula, matrix, lm and comparison bootstrap
PASS: two reduced contrasts retain their directions with rebuilt transformations and interactions
PASS: paper is the default and selected points are the CI statistics
PASS: nuisance formula and full matrix reproduce independent conditional moments
PASS: nuisance and cluster columns participate in the common case selection
PASS: cluster sampling preserves whole clusters and refits nuisance for every draw
PASS: nuisance rank loss is diagnosed without replacing bootstrap draws
PASS: nuisance transforms are rebuilt and explicit row ordering is enforced
PASS: conditional encoded geometry governs F_H after nuisance adjustment
FAIL: paper saturation is supported and nuisance temperature zero stays singular: 1 string mismatch
PASS: reference mean uses N degrees of freedom and preserves its direction
PASS: reference null and free mean share measurement and numerical F components
PASS: reference formula and matrix use uncentered encoded geometry
PASS: translation and paired sign preserve magnitudes without hiding direction
PASS: explicit nuisance retains only the declared reference space
PASS: explicit affine L beta=c matches independent constrained least squares
PASS: fixed-reference case draws carry row-specific origins and exact percentiles
PASS: reference cluster draws recompute N and resample the origin vector
PASS: reference comparison rejects changed origins and implicit centered partners
PASS: paired differences are recomputed on the supplied pair coding
PASS: paired selection keeps complete persons and response units
PASS: reference zero-signal and zero-temperature cases keep their meaning
PASS: reference and centered saturated ranks have their appropriate dimensions
PASS: reference response rescaling changes temperatures and F but not noncentrality
> cat(sprintf("\nLambda Star: %d passed, %d failed.\n", .count, length(.failures)))
Lambda Star: 138 passed, 1 failed.
> if (length(.failures)) stop(paste(.failures, collapse = "\n"))
Error: paper saturation is supported and nuisance temperature zero stays singular: 1 string mismatch
Execution halted
Flavor: r-release-macos-arm64
Version: 0.8.1
Check: tests
Result: ERROR
Running ‘core.R’ [15s/15s]
Running the tests in ‘tests/core.R’ failed.
Complete output:
> library(lambdastar)
>
> # Deliberately base R: the tests need no optional testing installation.
> .count <- 0L
> .failures <- character()
> test <- function(name, code) {
+ tryCatch({ force(code); .count <<- .count + 1L; cat("PASS:", name, "\n") },
+ error = function(e) {
+ .failures <<- c(.failures, paste(name, conditionMessage(e), sep = ": "))
+ cat("FAIL:", tail(.failures, 1), "\n")
+ })
+ }
> equal <- function(actual, expected, tol = 1e-10) {
+ result <- all.equal(actual, expected, tolerance = tol, check.attributes = TRUE)
+ if (!isTRUE(result)) stop(paste(result, collapse = "; "))
+ }
> relative <- function(actual, expected, tol = 1e-8) {
+ stopifnot(length(actual) == length(expected), all(is.finite(actual)),
+ all(is.finite(expected)), all(expected != 0),
+ all(abs(actual / expected - 1) <= tol))
+ }
> fails <- function(code, pattern) {
+ err <- tryCatch({ force(code); NULL }, error = identity)
+ if (is.null(err) || !grepl(pattern, conditionMessage(err)))
+ stop("Expected error matching: ", pattern)
+ }
> test_root <- if (dir.exists("cases")) "." else "tests"
> fixture <- function(name) file.path(test_root, "fixtures", name)
> for (file in list.files(file.path(test_root, "cases"), full.names = TRUE)) source(file)
PASS: measurement API preserves scale and case provenance
PASS: matrix, formula, lm and H1 paths agree
PASS: two-group adapter agrees with pooled-variance t-test
PASS: ANOVA, ANCOVA and interaction use the full encoded model
PASS: recoding preserves projection but can change geometry and F_H
PASS: lm stored contrasts survive changes to global contrast options
PASS: response mismatches and external formula variables are rejected
PASS: lm validates saved cases, design and response against common data
PASS: weights, offsets and other model families are rejected
PASS: rank deficiency, lost levels and invalid explicit row identity fail
PASS: mean response and F_H temperature stay coupled under item scaling
PASS: null model and temperature singularity remain explicit
PASS: required data, indicators, resolution and settings are validated
PASS: S3 summaries and printing expose estimates without full matrices
PASS: bootstrap defaults and explicit overrides are resolved and validated
PASS: same seed gives identical joint draws and changing seed changes draws
PASS: bootstrap preserves RNG kind and present or absent caller seed
PASS: joint resampling matches independent pairwise moments and lm.fit
PASS: percentile intervals use requested level and joint kappa ratios
PASS: missingness and strata use one original selection and retain row mapping
PASS: fixed-mixture variance requires estimable within-stratum variances
PASS: comparison shares draws with separate matrix/formula/lm pipelines
PASS: data-dependent transformations rebuild while factor coding stays fixed
PASS: lost factor levels fail only the affected model and never get replaced
PASS: zero temperature skips every CI including temperature without sampling
PASS: zero total structure never yields an invented kappa interval
PASS: item scaling transforms both point and interval noncentralities
PASS: retention is optional, thresholds explicit, summaries include diagnostics
PASS: numerical eta agrees with independent 90-digit Decimal reference
PASS: tiny eta follows analytic three-bin limit instead of 0.5
PASS: capacity scaling, vectorization and high-resolution behavior
PASS: thermal boundaries and smooth small-temperature stability
PASS: direct quadrature agrees with known primitive and small-ratio limit
PASS: lookup agrees with direct integration across range and off-grid
PASS: numerical controls are validated and limits fail explicitly
PASS: comparison equals separate evaluations and obeys swap/self identities
PASS: common missingness selection precedes both data-dependent designs
PASS: matrix missingness joins the same common complete-case selection
PASS: lm case omissions are accepted only when they match the common rows
PASS: two explicit models and a common matrix are required
PASS: comparison retains each formula coding and common numerical settings
PASS: transform-created missing and nonfinite values never alter cases silently
PASS: E3 preserves 0.5.0 whole and conditional adaptive-basis outputs
PASS: general multirow L agrees with independently fitted restricted models
PASS: coefficient, matrix and lm sources retain the same conditional geometry
PASS: full coefficient restrictions reduce to a whole-model state
PASS: L units and invertible recoding preserve capture but retain F_H geometry
PASS: L input errors and nonidentifiability are explicit
PASS: derived null and target are rebuilt for every complete-case bootstrap draw
PASS: contrast cluster draws preserve all rows and variable sample sizes
PASS: fixed and reevaluated polynomial/spline blocks reproduce their declared draw estimators
PASS: fixed adaptive nuisance and whole-model encoding are learned after selection
PASS: simple and explicitly averaged grid effects agree with named coefficient L
PASS: grid prediction uses learned poly, spline and scale coordinates
PASS: fixed grid contrasts use original learned basis with draw-specific restricted fits
PASS: fixed named polynomial coefficients differ from reevaluated coordinates as declared
PASS: grid validation prevents implicit weighting and unsupported prediction recipes
PASS: Fourier and explicitly fixed-knot piecewise/spline recipes equal encoded matrices
PASS: rank failures remain original draw failures without replacement
PASS: multirow grid recoding shares one null reference and retains factor coding
PASS: contrast singular temperature and undefined kappa retain their meanings
PASS: parallel measurements agree with independent pairwise identities
PASS: two-replicate identity and default composite scale
PASS: partition identity, scaling, and preserved observation identity
PASS: zero bath is singular while zero signal has undefined kappa
PASS: invalid measurements and row mismatches are rejected
PASS: tiny positive temperature is not clipped to zero
PASS: 0.4.1 baseline points, geometry, draws and intervals remain compatible
PASS: saved designs retain coding independently of global contrasts and caller functions
PASS: lm and matrix design objects preserve their encoded models
PASS: bound designs reject stale predictors and row order but permit new measurements
PASS: conditional blocks match restricted fits and actual encoded geometry on unequal cells
PASS: block transformations are rebuilt after shared selection and inside each draw
PASS: additional nuisance and conditional cluster bootstrap keep common row provenance
PASS: comparison requires explicit compatible references and preserves paired draws
PASS: conditional comparisons share nuisance while retaining distinct target columns
Tracing function ".nuisance_space" in package "namespace:lambdastar"
Tracing function ".thermal_terms" in package "namespace:lambdastar"
Untracing function ".nuisance_space" in package
"namespace:lambdastar"
Untracing function ".thermal_terms" in package "namespace:lambdastar"
PASS: shared point evaluation prepares nuisance and thermal terms only once
PASS: term selections, omissions and rank failures have explicit meaning
PASS: designs preserve singular temperatures and raw coding sensitivity
PASS: panel time trend matches explicit person dummy least squares
PASS: default panel occasion effect is categorical with literal coding
PASS: additional nuisance is projected within persons with correct rank
PASS: panel target blocks express group by occasion effects after person adjustment
PASS: person slopes rebuild separate copies in every bootstrap draw
PASS: unbalanced panel bootstrap retains each selected person's entire pattern
PASS: person resampling keeps stratum counts and rejects time-varying strata
PASS: panel complete selection removes whole persons on common comparison variables
PASS: panel comparisons reuse the same measurement and standalone numerical F
PASS: person fixed effects can be targets with distinct bootstrap identities
PASS: paired two-occasion reduction agrees with E6a lambda values and scaled F
PASS: panel scale and person-constant shifts preserve appropriate quantities
PASS: panel validation rejects ambiguous maps and incompatible estimator paths
PASS: panel fixed non-person bases and explicit matrices remain usable
PASS: declared person factor contrasts survive panel preparation and resampling
PASS: panel zero temperature and zero signal retain explicit diagnostic meanings
PASS: saturated person by occasion interactions retain positive bath df
PASS: unavailable library percentiles return diagnostics without substituted bounds
PASS: single-model F_H agrees with independent lm and determinant
PASS: recoding changes F_H solely through geometry
PASS: singular inference is distinct from the mathematical boundary
PASS: null model, smooth opt-in, and response-scale covariance
PASS: population estimator is independent of bootstrap and confidence level
PASS: every stored draw matches independent pairwise moments and OLS
PASS: percentile construction agrees with boot on the identical person draws
PASS: weak signal records undefined ratios without substituting an interval
PASS: valid zero/one ratios and degenerate distributions retain their percentiles
PASS: fixed-stratum population draws match independent mixture variance
PASS: null models have valid zero localization for positive signal
PASS: saturated models keep finite-person F_H and identify unestimable population projection
PASS: singular temperature draws preserve temperature zeros and report dependent failures
PASS: zero total point retains a lambda interval and an undefined kappa point
PASS: unrepresentable tail probabilities are reported without substituting limits
PASS: known orthogonal energy partition is 9 plus 16
PASS: encoded geometry differs while recoded projection agrees
PASS: null, one direction, rank loss, and bad projection
PASS: reduced factor contrasts agree across formula, matrix, lm and comparison bootstrap
PASS: two reduced contrasts retain their directions with rebuilt transformations and interactions
PASS: paper is the default and selected points are the CI statistics
PASS: nuisance formula and full matrix reproduce independent conditional moments
PASS: nuisance and cluster columns participate in the common case selection
PASS: cluster sampling preserves whole clusters and refits nuisance for every draw
PASS: nuisance rank loss is diagnosed without replacing bootstrap draws
PASS: nuisance transforms are rebuilt and explicit row ordering is enforced
PASS: conditional encoded geometry governs F_H after nuisance adjustment
FAIL: paper saturation is supported and nuisance temperature zero stays singular: 1 string mismatch
PASS: reference mean uses N degrees of freedom and preserves its direction
PASS: reference null and free mean share measurement and numerical F components
PASS: reference formula and matrix use uncentered encoded geometry
PASS: translation and paired sign preserve magnitudes without hiding direction
PASS: explicit nuisance retains only the declared reference space
PASS: explicit affine L beta=c matches independent constrained least squares
PASS: fixed-reference case draws carry row-specific origins and exact percentiles
PASS: reference cluster draws recompute N and resample the origin vector
PASS: reference comparison rejects changed origins and implicit centered partners
PASS: paired differences are recomputed on the supplied pair coding
PASS: paired selection keeps complete persons and response units
PASS: reference zero-signal and zero-temperature cases keep their meaning
PASS: reference and centered saturated ranks have their appropriate dimensions
PASS: reference response rescaling changes temperatures and F but not noncentrality
> cat(sprintf("\nLambda Star: %d passed, %d failed.\n", .count, length(.failures)))
Lambda Star: 138 passed, 1 failed.
> if (length(.failures)) stop(paste(.failures, collapse = "\n"))
Error: paper saturation is supported and nuisance temperature zero stays singular: 1 string mismatch
Execution halted
Flavor: r-oldrel-macos-arm64