Refine demangler setting and execinfo use#1445
Conversation
kevinushey
left a comment
There was a problem hiding this comment.
LGTM!
Sorry, something went wrong.
b96082e
into
master
Jan 13, 2026
|
FWIW and sorry to be a nuisance, I had suggested a KISS solution in #1400 (the "whitelist" approach, where I did mention that "BSDs have long provided execinfo, too, but a web search suggests they require separate linking with -lexecinfo"). That would be much simpler than the now-restored bunch of conditionals (including the non-standard |
Sorry, something went wrong.
|
Thanks. Might it be worthwhile to add the hint about linking needed to the short paragraph in R-exts.texi?
|
Sorry, something went wrong.
Obtained from: RcppCore/Rcpp#1445 Reported by: portscout
Obtained from: RcppCore/Rcpp#1445 Reported by: portscout
PR #1401 updated the detection for
execinfo.hby relying on a header-finding macro directly which is powerful. It also removed a set of#if defined(...)tests precluding several architectures from trying. This now bit us in issue #1442 as on FreeBSD, the header is found ... but also requires a library.Now, we could try to use autoconf to find header and library, and attempt to compile a test program (as we now do for OpenMP in RcppArmadillo and RcppEigen). But given that the affected architectures here are on the less-widely used side with difficult access for us to test and refine, a simpler strategy would be to punt and again rely on a set of
#if define()so this PR restores this. That is plain KISS Principle: Keep it Simple (and) Stupid.Checklist
R CMD checkstill passes all tests