Aaron Richton wrote:
Which test were you running?
Do you have the tests/testrun directory? If so, preserve a tarball of that in case it's needed.
Run "threads" in dbx, and do a "where" in each thread:
Segmentation fault in noname.strlen [/usr/lib/libs.a] at 0xd0335680 ($t14) 0xd0335680 (strlen) 89030000 lbz r8,0x0(r3)
Strictly speaking, I can't rule out a bug in AIX's strlen(), but the odds that that's what you want us looking at are infinitesimally small.
Was running test050 500 times. Been watching threads on test038 and test050 and decided to stress them on AIX as well. I am betting there is an intermittent pointer corruption issue that periodically triggers a segmentation fault when calculating the string length. However, I was ill-prepared to provide the debug detail of which I knew everyone would be interested. :-) :-)
Cheers, Bill