Regression testing buildworld on RPI3

[bob prohaska]

After much trial by error, it's looking as if -j4 buildworld runs 
without error on r339376 and stops with signal 11 in cc somewhere
near or after r342204. 

Can anybody suggest which individual revisions should be tested
to locate a point of failure? Buildworld on a Pi3 takes about 
a day. Recovering from a failure, if it gets into userland, is 
much slower. I gather the problem is not apparent during a cross-
compile on AMD64.

The plan for now is to update, compile the kernel toolchain and
a new kernel, then reboot and try to build world.  IIUC this will 
exercise the new world without disturbing userland. Is this correct?
If there's a failure, I can reboot the old kernel, back down the
revision, and try again.

Thanks for reading, and any thoughts.

bob prohaska