[R-pkg-devel] Possible gcc-10.0.1 Bug in gcc-10 Test Server

[brodie gaslam]

At the risk of getting laughed out of the room I wanted to share with you some interesting findings that arose from trying to figure out why my package vetr is segfaulting on the gcc-10 test machine:

https://www.stats.ox.ac.uk/pub/bdr/gcc10/vetr.out

Thanks to Dirk Eddelbuettel who pointed me to rocker/r-base that builds on the testing debian repo I was able to reproduce the error, and figure just enough to post a semi-cogent question on SO:

https://stackoverflow.com/q/60406042/2725969

And some folks that know more than me about compilers rendered judgment that it likely is a compiler error.  I realize "random dudes on SO rendered judgment" is not exactly the most ringing endorsement, but the logic they present is sound and matches my reading of the C standard (not that that is worth much).

It is unlikely for others to run into this issue as it requires an extremely particular set of circumstances.  If you too segfault on Professor Ripley's machine please don't just point to this as the reason without actually confirming that's what's happening.

Best,

Brodie.