The tests are currently failing:

══ Failed tests ════════════════════════════════════════════════════════════════
── Failure ('test-unvalidated-summary.R:90:3'): summary.simtrial_gs_wlr() returns consistent results for one-sided design ──
`observed` not equal to `expected`.
Component "asy_upper_prob": Mean relative difference: 4.487592e-07
Component "asy_n": Mean relative difference: 8.453473e-07
Component "asy_event": Mean relative difference: 8.453473e-07
── Failure ('test-unvalidated-summary.R:182:3'): summary.simtrial_gs_wlr() returns consistent results for two-sided design ──
`observed` not equal to `expected`.
Component "asy_upper_prob": Mean relative difference: 4.137911e-07
Component "asy_n": Mean relative difference: 8.823477e-07
Component "asy_event": Mean relative difference: 8.823477e-07

[ FAIL 2 | WARN 0 | SKIP 0 | PASS 254 ]

In PR #324, we started installing the dev version of {gsDesign2} to obtain the latest features.

In PR #326, I made a trivial change. The CI passed for the PR but failed for the merge commit b41182c

I was able to reproduce this locally. The tests in test-unvalidated-summary.R fail with {gsDesign2} installed from GitHub but pass when {gsDesign2} 1.1.3 is installed from CRAN.

Based on the timing of the test failure, I suspect the cause was the merging of PR Merck/gsDesign2#528

Note that this is a great motivating example of why I wish we would squash and commit PRs (especially large ones). Merck/gsDesign2#528 included 87 commits made over the course of weeks. If it was squashed into a single commit, it would have been easy for me to confirm if it was the cause.

Are we concerned about these unexpected test failures? I can fix them by setting tolerance = 1e-6, but I wanted to confirm this was an acceptable tolerance first.