general protection fault in 1.10

[jbr...@easypost.com]

This is a weird one. We run a few beanstalkd servers in production, and one of them just died with a general protection fault. Nothing in beanstalk's log, but got the following in dmesg:

traps: beanstalkd[17694] general protection ip:401e16 sp:7fffe98761c8 error:0 in beanstalkd[400000+e000]

No interesting changes in memory use or anything else.

I'm tempted to write this off as a cosmic ray flipping a bit in EIP or something like that, but I figured I'd send it to the list in case anyone else had seen any such weird behavior and there's some subtle bug in the code.

This is v1.10 built as an RPM (by us) for CentOS 6 under gcc 4.4.7, with the patches from https://github.com/kr/beanstalkd/pull/218, https://github.com/kr/beanstalkd/issues/276, and https://github.com/kr/beanstalkd/pull/296 applied.

[jbr...@easypost.com]

Just got another segfault, on a different machine, and in a different place. Details for our most recent debugging session are at https://github.com/kr/beanstalkd/issues/328 if anyone has any thoughts.

[Urjit Singh Bhatia]

We just hit this in production today! After running for an year without issues. As you pointed out, it did happen when I was removing some connections...