Closed eddelbuettel closed 1 year ago
Rather than commenting out the failure, I'm giong to reinstate those tests to include 1:5
, but change the expected outcome.
I guess there has been an update to the 'spline' algorithm(s) in the latest boost? Because the error is only changing the value of the 'green' channel by one value
#80800064
to #807F0064
Which I'm happy to live with (and is probably imperceptible to humans?)
Fully agreed. Five is better. And I have at times adjusted Rout.save to match newer Rout too, back in the day.
Hi Team SymbolixAU,
We are in the annual BH upgrade process right now. Boost updates every four months, we skip the April and August updates and try to follow each December with one annual update. Boost, and hence BH, 1.81.0-0 looks good: only a handful of packages were affected when we did reverse depends checking, and two of those were actually affected by the switch to
g++-12
we made at the same time. See https://github.com/eddelbuettel/bh/issues/88 for the full details, and example commands to install thisBH_1.81.0-0
release from a drat repo, or r-universe. We hope to update at CRAN in mid-January,colourvalues
looks good too. The only noticeable change were two tests comparing five colour values each differing on one in each comparison. The first of the two commits here simply takes those differing colours out of the comparison.The second commit simply switches from C++11 to C++14 which is the default under R 4.2.* anyway. In doing so we avoid a nag message from Boost telling us that C++14 will be the required minimum for Boost.Math come Boost 1.82 in April.
I hope you can apply this PR which tests cleanly on my test system, and get an updated
colourvalues
to CRAN in early January. Thanks in advance for your consideration of this.