On Wed, 9 Sep 2020 13:33:05 -0600 Jerry James loganjerry@gmail.com wrote:
On Wed, Sep 9, 2020 at 1:12 PM Ranjan Maitra maitra@email.com wrote:
Thanks, so this appears to be a Fedora problem. Btw, I tried oracle's jdk-14.0.2_linux-x64_bin.rpm and had the same result.
I filed a bug report: https://bugzilla.redhat.com/show_bug.cgi?id=1877524
Why does everyone seem to think this is a Java bug? The system log showed that a crash occurred in Rf_eval (libR.so + 0x157ab3); i.e., in the R library. The stack trace shows a lot of calls to these functions:
bcEval forcePromise getvar R_execClosure Rf_appyClosure Rf_eval
It's possible that R is being handed an expression to evaluate that is so complex that it runs out of stack space, or there could be an infinite recursion bug in R somewhere. If the former, the solution is to bump up the stack space. I would try the -Xss argument to java first to see if that makes the issue go away. I'm not sure what the default value is, so I don't know what to suggest, but maybe start with something like -Xss4M. If that fixes the issue, then retry with successively smaller values to find something reasonable that is still big enough to work.
Jerry,
Thanks for pointing this out. So, on the commandline for R, the equivalent call:
ggplot() + geom_boxplot(aes(y = speed), data = cars)
plots without a hitch. Now boxplots require calculations of three quartiles, and a couple more calculations to decide on outliers, and this may be a R problem, but this is a 50-observations dataset. Besides, the call inside JGR works with Windows as well as Mac and on Ubuntu 20.04 (per George), but crashes things on Fedora.
How do I add the argument to java that you suggest? Where do I do this?
Many thanks, Ranjan