eCos bitrot

[Alex Schuilenburg <alexs@ecoscentric.com>]

Hi Dudes

The eCosCentric test farm is making pretty good progress, most of it is 
not visible yet unfortunately (yes, my fault, I am kept occupied on 
other trivial but necessary stuff).

Anyway, here is a brief update:
* We are building perms and tests for various targets
* We are running automated tests for those targets
* Results are in raw log files which are easily turned into HTML,
   including summaries, although we are not wasting time tidying this as
   we are going for the final result: logging the results in the
   database.

The results so far for anoncvs have been very interesting:

Builds
------
*  1 perm now fails to configure
* 11 perms will no longer build
*  3 perms produce errors building tests (make tests error)
* *all* perms produce build warnings

Tests
-----
So far we have only put in the EBSA targets (courtesy ASCOM) as the 
other targets need work: For e.g. PID:- Need RedBoot in EPROM and we 
don't have a suitable programmer; PLC2 and LAKI:- need to get platform 
specific code and permission from ASCOM;

The EBSA has around a 90-95% pass rate for the tests that actually 
build, which is not bad considering.  All are timeouts (cache1, kcache1, 
etc). We (eCosCentric) are looking into these. I hope to have the build 
and test result display working within a week or so.  We can make the 
raw results available if anyone wants to jump the gun and fix the 
problems before the result pages are displayed - just let me know.

Jld is putting in some additional targets this afternoon and we should 
be up to 5-10% of the size the farm was at Red Hat shortly :-)  We need 
to extract the serial port terminal server from Bart and get that 
working to get more targets in place, although we are no longer 
restricted to requiring a serial port to determine reset OK.


Summary
-------
On the whole, we are pleased with the stability of the new farm. 
Results appear a lot more consistent (i.e. less noise and 
spurious/WFM-style errors) and we managed to run 4854 tests in a 24 hour 
period for a single platform.  So a full cycle should take around 3-4 
days insead of the usual 2 weeks.

We hope this will be a very useful resource for you and an extremely 
beneficial service for our customers.

Paul is the only one complaining as the rats have moved out the test 
farm and into his house because of the constant clacking ;-)

-- Alex