tests/testthat/test-nlmsetup-fresh-rx.R

In nlmixr2est: Nonlinear Mixed Effects Models in Population PK/PD, Estimation Routines

nmTest({
  test_that("nlmSetup does not segfault on first estimator call in a fresh R session", {
    # Regression test for nlm.cpp where
    #   nlmOp.stickyRecalcN2Per.assign((size_t)getRxNsub(rx), 0)
    # was called BEFORE rxode2::rxSolve_() initializes the global `rx`
    # pointer.  In a fresh R session `rx` is still NULL and the read
    # inside getRxNsub() dereferences a NULL pointer ("memory not mapped,
    # address 0x10").
    #
    # Affected every pooled estimator that goes through .nlmSetupEnv:
    #   bobyqa, nlm, optim, nls, nlminb, lbfgsb3c, n1qn1, newuoa, uobyqa.
    # focei uses foceiSetup_ and was unaffected.
    #
    # The regression only fires when `rx` has not yet been populated by
    # an earlier rxode2 solve or focei fit, so this test MUST run in a
    # truly fresh R session -- running inline in the testthat process is
    # not enough because earlier tests will already have populated `rx`
    # (per Matt Fidler's review on #653).  callr spawns a clean
    # subprocess; if the segfault returns the child dies and
    # callr::r() raises a callr_status_error which surfaces as a test
    # failure.  stats::nlm keeps the test cheap; the crash is in setup
    # so iterlim = 1L exercises the regression path.
    skip_on_cran()
    skip_if_not_installed("callr")

    cls <- callr::r(
      function() {
        library(nlmixr2est)
        mod <- function() {
          ini({
            tka <- 0.45
            tcl <- log(2.7)
            tv <- 3.45
            add.sd <- 0.7
          })
          model({
            ka <- exp(tka)
            cl <- exp(tcl)
            v <- exp(tv)
            linCmt() ~ add(add.sd)
          })
        }
        fit <- nlmixr2(
          mod, nlmixr2data::theo_sd, est = "nlm",
          control = nlmControl(iterlim = 1L, print.level = 0L)
        )
        class(fit)
      }
    )
    expect_true("nlmixr2FitCore" %in% cls)
  })
})