On Sun, 4 Sep 2011, Henrik Grubbstr?m (Lysator) @ Pike (-) developers forum wrote:
According to git bisect (and your testcase), the bug was introduced in commit 90314251db074fc740a2a2b01a9aca1939cd91a4.
Thanks for bisecting. The reason why my change triggered this one is that I somehow overlooked the fact that the previous hash size initialization had a minimum. I fixed that now.
It looks like your "fix" broke the testsuite:
| $ ./pike -mmaster.pike -x test_pike testsuite | Forked watchdog pid 21681. | Begin tests at Mon Sep 5 19:40:13 2011 (pid 21680) | Doing tests in testsuite (11333 tests) | /home/grubba/src/Pike/7.9/src/mapping.c:2366: Fatal error: | Pretty mean hashtable there buster 32 > 11 (2)! | Pretty mean hashtable there buster 32 > 11 (2)! | Backtrace at time of fatal: | testsuite:1: testsuite()->a() | /home/grubba/src/Pike/7.9/lib/modules/Tools.pmod/Standalone.pmod/test_pike.pike:1059: | Tools.Standalone.test_pike()->main(2,({"/home/grubba/src/Pike/7.9/lib/modu | les/Tools.pmod/Standalone.pmod/test_pike.pike","testsuite"})) | Abort (core dumped)
If I revert the "fix", it works again. It looks like the complaint is that there are too few keypairs allocated compared to the size of the hashtable.