On Thu, Oct 8, 2009 at 11:00 PM, openbsd.misc.tmp openbsd.misc.tmp <[email protected]> wrote: >> > What I have is application.core, but I cannot read this: >> > "/home/me/application.core " is not a core dump: File format not recognized >> > (gdb) quit >> >> did you check the core size to see if it didn't get truncated >> because of some ulimit ? > > Yes, the core file is definitely a lot smaller than what ulimit -c was set to. > > Furthermore the dump should have a header so that 'file' could > determine the file type. Which is not the case.
First of all, 4.3 is no longer supported, so if there's an actual bug you're on your own. 4.6 is about to be released so you should be long into planning your upgrade process. Also, you failed the first step of a bug report by leaving out your dmesg. > This happened more than once. Yet, all the .core files I was able to get > were unusable to me. However, when I initiate a .core file by exiting the > application intentionally with a signal, the resulting core files are OK. So you have evidence that the kernel generates good core files. That suggests the hypothesis that these files aren't actually generated by the kernel. Do you have evidence that they are? Log statements somewhere? A nanny process that got the abnormal exit status from wait()? Sounds like it's time to crank up the logging around the application, write a 'nanny' script that wraps the process and dumps information about its death when it exits, or run the thing under gdb to start with. You face a mystery: time to crank up the scientific method! Make a hypothesis, figure out how to test it, run the experiment, and draw a conclusion. If you're going to stay with 4.3 then that will be your only option... Philip Guenther
