fix ATLAS issue

DESCRIPTION

@@ -1,7 +1,7 @@
 Package: parglm
 Type: Package
 Title: Parallel GLM
-Version: 0.1.5
+Version: 0.1.6
 Authors@R: c(
   person("Benjamin", "Christoffersen", 
          email = "[email protected]", 

cran-comments.md

@@ -1,27 +1,16 @@
 ## Test environments
 * Ubuntu 18.04 LTS with gcc 8.3.0
   R version 3.6.3
-* Ubuntu 18.04 LTS with gcc 8.3.0 with `--use-valgrind`
-  R version 3.6.3
 * Ubuntu 16.04 LTS (on travis-ci)
   R version 4.0.0
-* Ubuntu 18.04 LTS with clang 6.0.0 with ASAN and 
-  UBSAN checks
-  R devel (2020-07-02 r78770)
 * win-builder (devel and release)
-* `rhub::check_on_solaris()`
 * `rhub::check_for_cran()`
-* Fedora 30 with openBLAS like on CRAN's issue kinds page.
 
 ## R CMD check results
 There is a note about the size of the package on some platforms.
 
-I have tried to reproduce the error with openBLAS and Solaris but failed 
-to do so. I made a post on R-pkg-devel 
-(https://stat.ethz.ch/pipermail/r-package-devel/2020q1/005175.html)
-but I did not get a reply.
-
-## Resubmission
-This is a resubmission. In this version I have:
-
- - made the tests more stable numerically.
+I changed one of the tests such that it no longer checks the number of 
+iterations used by the method. This is the reason the test failed on ATLAS 
+which caused the package to get pulled of CRAN. However, the estimates 
+where similar to the test values and hence checking the iteration number 
+makes little sense.

tests/testthat/test_parglm_fams.R

@@ -240,7 +240,7 @@ test_that("'method' equal to 'LINPACK' behaves as 'glm'", {
   s2 <- summary(f2)
 
   excl <- c("call", "coefficients", "cov.unscaled", "cov.scaled",
-            "dispersion")
+            "dispersion", "iter")
   expect_equal(s1[!names(s1) %in% excl], s2[!names(s2) %in% excl])
   expect_equal(s1$coefficients, s2$coefficients)