:: Re: [devuan-dev] Jenkins udisks2/ar…
Αρχική Σελίδα
Delete this message
Reply to this message
Συντάκτης: Andreas Messer
Ημερομηνία:  
Προς: devuan-dev
Αντικείμενο: Re: [devuan-dev] Jenkins udisks2/arm64
Mark,

On Tue, Oct 15, 2019 at 09:02:47PM +0100, Mark Hindley wrote:
> Andreas,
>
> On Tue, Oct 15, 2019 at 09:28:10PM +0200, Andreas Messer wrote:
> > Hi,
> >
> > I somehow remind a mail from Leepen containing something
> > about a build issue on Jenkins with udisks2. I have investigated
> > a little and seen that the last build on 2nd May 2018 has still
> > failed. Since there has already passed quite some time since then,
> > I wonder if this is still an issue?
>
> The arm64 build isses I reported were with policykit-1. rrq also had the
> same failure (oom when compressing lzma) with util-linux. I haven't even
> tried a build of the new udisks2 until this is fixed.o


Thx for correcting me. I just had some faint memory about it.

> If you have any insight into why the arm64 buildhost is running out of
> memory, please do say.


I had a look at the policykit build. So AFAIK there are two, distinct issues:

- Bus Error when testing udisks2 package
- Out of memory when compressing the archives of different packages

The later seems strange to me. I don't know much about the build nodes of
the CI system neither about the used flow (the build script / Jenkins
settings). But assuming that the hardware for the arm64 build is an
Overdrive 1000 machine which is/was shipped with minimum of 8GB RAM,
I would expect no such problem at all. I'm managing a Jenkins Build /
Hardware-In-The-Loop test - system at our company which is running roughly
more than 30 gcc/ld/ar/python/zip/whatever instances in parallel during build
and never observed a memory shortage. (Also 8GB RAM in that VM instance)

I think to find & tackle this issue, direct access to the machine & knowledge
about the configuration is necessary. Probably only parazyd can do this.

The former issue should be reproducible on my Librem 5 devkit without
further knowledge. I could examine this, but not before this weekend.

Cheers,
Andreas

--
gnuPG keyid: 8C2BAF51
fingerprint: 28EE 8438 E688 D992 3661 C753 90B3 BAAA 8C2B AF51