Re: gpsd | Pipeline #200139571 has failed for master | 71f2b070

[Gary E. Miller]

Yo Bernd!

On Thu, 15 Oct 2020 21:44:33 +0200
Bernd Zeimetz <bernd@bzed.de> wrote:

> On 10/14/20 8:56 PM, Gary E. Miller wrote:
> > Almost of the CI time appears to be just building the test image,
> > not the tests.  Fix that big problem, not the problem you want to
> > fix.  
> 
> Appears?
> Please prove it.

Check my first email on this topic.

> > Which fixed persistent FreeBSD failures.  There is some serious flaw
> > in the FreeBSD kernel.  Slow, or correct, choose one.  
> 
> Nobody every figured it out if the issue is in the kernel or in the
> gpsd code. Or both. Or in Python. Or libc.

Of course we do not know where the problem is, toerhwise it would be
fixed. But is sure seems to vary by kernel type (FreeBAD, Linux), so it
sure feels like a kernel problem to me.  Feel free to show me where the
problem is.

> >> Nothing that is fixable with CPU power or whatever else.
> >> Just by fixing the crap that needs WRITE_PAD at all.  
> > 
> > You, and everyone else, wants that.  Care to take a stab at it?  
> 
> esr spent various hours on it and so I've done.

As have I and many others.  Not an easy problem.

> If you have a trivial and reproducible way to trigger the issue,
> please show me. So far it happens just random.

Easy, just reduce WRITE_PAD on FreeBAD.  Fails every time in the CI.

> Don't forget that we had write_pad for linux for some time before. Its
> not like this is a freebsd only problem, maybe it only works in linux
> as the cpu speed got faster these days.

Dunno.  We don't know until we know.

RGDS
GARY
---------------------------------------------------------------------------
Gary E. Miller Rellim 109 NW Wilmington Ave., Suite E, Bend, OR 97703
        gem@rellim.com  Tel:+1 541 382 8588

            Veritas liberabit vos. -- Quid est veritas?
    "If you can't measure it, you can't improve it." - Lord Kelvin

pgpGz5XbwAZUT.pgp
Description: OpenPGP digital signature