making sense of crashes in build logs

[David Holmes]

On 7/04/2024 1:52 am, Thorsten Glaser wrote:
> Hi,
> 
> how do I make sense of crashes in buildd logs, such as:
> https://buildd.debian.org/status/fetch.php?pkg=openjdk-8&arch=m68k&ver=8u402-ga-6&stamp=1711506075&raw=1
> 
> I saw…
> #  Internal Error (os_linux_zero.cpp:254), pid=830846, tid=0x3f629440
> … and thought, yay, we can look at where the segfault is,
> but that address is the generic signal handler ☹
> 
> Is there a way to get a backtrace into these messages?

If there is no backtrace in the hs_err log then it is because the VM was 
not in a state where stack walking (as per the port) was possible. If 
there was a core file then you may be able to debug that. Otherwise if 
this crashed during the build you should be able to run the crashing JVM 
directly and potentially debug things further that way i.e.

 > configure:5339: checking if the VM and compiler work together

Try running the underlying command directly.

Is this running under an emulator? The build version is for m68k but the 
crash log shows an x86 machine.

David
-----

> Building locally in a VM would take… over a week, I think,
> and gdb is also currently not very usable on m68k, at least
> it wasn’t when I was trying to debug a different crash in a
> different software…
> 
> Help with m68k is welcome in general; the last version that
> built was 8u171-b11, the first to fail was 8u181-b13 and all
> others since. Some patch probably needs updating or something.
> 
> Thanks,
> //mirabilos