Re: next version of Lout

[Mark Summerfield]

On 2008-09-28, address@hidden wrote:
> Jeff wrote:
>  > I'm hoping that someone will find out why
>  > prg2lout is crashing under Ubuntu between now and then.
>  > I did read carefully through prg2lout.c and change a
>  > couple of things that looked slightly suspicious, but
>  > Mark Summerfield reported that the changes did no good.
>  > So I've basically done everything I can do here without
>  > a core dump to guide me.
>
> Our local Ubuntu developer and guru explained to me:
>  | Ubuntu has a special program ("apport") to detect crashes and report
>  | information about them (including stack backtraces) to the Ubuntu issue
>  | tracker.
>  |
>  | The kernel has a sysctl variable to support this called
>  |
>  |   kernel.core_pattern
>  |
>  | if set to a string starting with '|' then it invokes the named program
>  | instead of writing the core to the named file.  The kernel default is
>  | 'core'.
>  |
>  | You can check with 'sysctl kernel.core_pattern'.
>  |
>  | Ubuntu sets this to redirect crashes to their special handler program.
>  |
>  | If you set it back (see /etc/sysctl.conf or "man sysctl") to "core" and
>  | set "ulimit -c unlimited" then you'll get core files in the current
>  | working directory when a program crashes.
>
> But I am not running Ubuntu myself...

Thanks... I've replaced every exit(1) call in prg2lout with abort() to
force core dumps and I just checked and syctl reports that
kernel.core_pattern = core. I've come to the view that prg2lout is
terminating and reporting 1 but not through anywhere it is supposed to
though I don't know how of course. Since this seems very obscure
(e.g., doesn't affect Fedora, and only affects Kubuntu in certain
circumstances), and since it is easy to work around, I've given up!

-- 
Mark Summerfield, Qtrac Ltd, www.qtrac.eu
    C++, Python, Qt, PyQt - training and consultancy
        "C++ GUI Programming with Qt 4" - ISBN 0132354160