MMMmm

[Syzop]

Jeremy Allison wrote:

> If you want something to be investigated, a good description,
> with a *REPRODUCIBLE TEST CASE* is needed.
>
> For example, I'm going to great efforts with the Win9x clarion
> problem that someone has reported (haven't been able to
> reproduce it yet).
>
> This is because the person included a link to the test code,
> a full description of the samba versions being used, on what
> platforms, and a full description of what the clients were
> doing.
>
> This gets attention. I vaguely remember your posts, so I
> can't say they didn't include all that, but you might want
> to ensure you have those details and re-post.

This is exactly what I tried, I had a log file of 1,5Gb, tried to
search in the source code where it went wrong, etc.
But indeed I can't give you a "reproducible test case",
at home I can't reproduce it, only at work with XX computers,
with many silly users who just turn the computer off in the
middle of whatever-program (which I suspect is the source of
the problem). I hope you want to re-read that post, and tell
me how I can help (I still have the 1,5Gb log). I also wanted
to change stuff in the source and see if it helps*, but it was a little
bit too risky because when things go wrong, many users can't work
so I wanted some "expert advise" at it.

*) Changing the behavior of failing when sending an oplock_break to
    a child from 'locking out that file forever' to 'removing the oplock'.

    Bram Matthys.