Hi, to keep you informed on the mips* buildd status: Prior to the breakage of ftp-master things looked pretty well. Mipsel had not only no backlog, but enough spare cycles to start building experimental packages. As of now, we have two hardware failures, so we have only one mipsel buildd left. We have three mips buildds, but two of them reboots themself if load gets too high (details not yet clear).
Our original plan was to convert one current mips(el) machine to the
other endianess as soon as we have enough new buildds (our old machines
are bi-endian, our new ones not – so we could convert one existing
machine in case we get mips or mipsel running stable to help the other
Currently mipsel mostly manages to avoid growing the queue, so every
package should be built “sometimes”. This mostly means that packages
need a few days to wait till they are finally built (new buildd sorting
order helps a lot here). Mips has an shrinking backlog, and will hit an
empty queue soon.
For mips, we have movidis boxes as new hardware. We basically have a
working kernel, with a few not-yet-upstream-patches. However, that
machine reboots itself sometimes. This could be triggered by load or
running irqbalance. [1, 2]
This doesn’t prevent us from using two such machines as buildd, and
we’re about to debug that together with upstream. Another machine of the
same set works as new porter box, gabrielli.
For mipsel, we also have a Longsoon 2F-machine (in addition to an
2E-machine), but any of the currently existing machines has the same
silicon level bug that allows user code to crash the machine
(machines without this bug are expected in summer). This makes it
hard to use it as buildd. Good news is that the cpu errata is now
 en: http://groups.google.com.hk/group/loongson-dev/msg/e0d2e220958f10a6?dmode=source
Also, we have two cases of hardware failures:
Mayr (mips) has a broken harddisk for some time. Our plans are to add
faster sata disks on this occasion, which however requires an additional
Rem (mipsel) refuses its current kernel, which however had already
worked on that machine.
Work is in progress to fix these two hardware issues.