[racket-dev] Building Racket with the Android NDK

From: Matthew Flatt (mflatt at cs.utah.edu)
Date: Wed Sep 5 09:18:24 EDT 2012

At Wed, 5 Sep 2012 12:59:39 +0100 (BST), Jan Wedekind wrote:
> The program now fails in 
> 'scheme_extend_module_rename_with_shared'. The stack trace has a couple of 
> hundred entries, so I am only posting part of it here. I hope the 
> backtrace makes any sense.

Nothing stands out, unfortunately. It's not even clear how the
identified source line could crash with a SIGSEGV, assuming that the
source report is accurate.

A strange crash with a moderately deep stack makes me wonder about the
stack size. Since the statically linked version works --- which
suggests that it's not a problem detecting the OS's stack size for the
main thread --- is threading any different with this dynamically linked
version? On some OSes, linking in a way that enables pthread support
causes main() to be called in a new thread, which might have a smaller
stack than the process's initial thread.

Posted on the dev mailing list.