Re: Bug# 1279147 - Gateway sigsegv's when cleaning up channels using ca_clear_channel

[Michael Davidsaver <mdavidsaver@bnl.gov>]

On 02/11/2014 09:08 PM, Shankar, Murali wrote:

...However, the place where this crashes is in a fundamental place (tsDLList.h:238) in EPICS base. What seems to be happening here is that we have an element in the linked list that has a previousNode of 0 but is itself not the pFirst element. I can provide more details/core(s) if needed.

Based on a quick check I think that locking is being done correctly.  So my guess is a list node handling problem (eg. double removal or removal from the wrong list).

I would suggesting adding some debugging code to tsDLNode to enable it to track which list it is current part of.  If list handling is the problem, then this should trap the error quickly.

This does not seem to be a gateway bug; it seems to be some issue in ca_clear_channel. However, I don’t want to change EPICS base; perhaps I can catch the exception in gatePv.cc:240 and then move on. Should I consider patching this like so in the gateway code? I know this has memory leaks but this does not happen often.

I don't it is possible to continue after SIGSEGV.