Bug#769864: new vcmi and fuzzylite package uploads

[Johannes Schauer]

Hi Stephen,

Quoting Stephen Kitt (2014-11-30 19:19:00)
> Indeed, I was a bit too optimistic, it still fails to build on mips and
> mipsel. But armhf, hurd-i386, kfreebsd-{amd64,i386} and ppc64el built fine,
> well done!

thanks for the bug handling!

I'm currently on to the mips/mipsel bugs.

With mips it might just be the buildd as vcmi built successfully in the past
but failed for the last upload which was done on a buildd with 1.8 GB of
memory:

https://db.debian.org/machines.cgi?host=lucatelli

instead of on one with 8 GB:

https://db.debian.org/machines.cgi?host=mips-aql-02

See the logs here:

https://buildd.debian.org/status/logs.php?pkg=vcmi&arch=mips


Unfortunately on mipsel, the build was always done on mayer with a meager 1 GB
of memory:

https://db.debian.org/machines.cgi?host=mayer

And here the log history:

https://buildd.debian.org/status/logs.php?pkg=vcmi&arch=mipsel


Since I'm not a DD and do not have access to a porterbox (and my request to get
access got no reply yet) I set up a mipsel qemu machine with 4 GB of RAM. This
was a bit tricky since at least according to:

https://people.debian.org/~aurel32/qemu/mipsel/README.txt

32-bit mipsel is limited to 256 MB of memory. So I put the qcow2 image in a
tmpfs on my host and created a 3 GB swap file in the emulated mipsel
environment. Since the whole qcow2 image is in RAM on my host, it should not
matter that the machine runs with 256 MB emulated RAM and 3 GB swap as it all
translated to system memory on my host.

Unfortunately the whole thing is super slow and already churning along for two
days and is only at 38% now.

I will report back with results of this experiments once I have them.

If the build finishes fine, we might have to approach buildd admins to restrict
the build of vcmi to a certain subset of mips and mipsel buildds.

cheers, josch