Re: [ECOS] Release cycle?

[Andrew Lunn <andrew at lunn dot ch>]

On Fri, Oct 28, 2005 at 08:40:26AM +1300, John Carter wrote:
> On Thu, 27 Oct 2005, Andrew Lunn wrote:
> 
> >eCos 2.0 is very old, something like 3 years.
> >
> >The problem you are having was fixed a couple of years ago. Update to
> >the latest code.
> 
> 
> Umm, this does bring to mind... What is the release cycle for eCos? It 
> would seem following the general drift of queries and answers of late 
> (it's been fixed, it's in CVS), that at very least a eCos version 2.1 is 
> needed.

> Traditionally one doesn't use code from the tip of CVS mainline in 
> production software because enhancements have destabilized it and 
> have introduced a rash of new bugs.

And where do these bugs get fixed? In the tip of CVS mainline.....

> Is the answer then that no new enhancements have gone in in the last three 
> years?

No, lots of enhancements and bug fixed have been added in the last
three years.
 
> One of the things that have always concerned me about eCos is the C bit. 
> Configurable. How on earth do you test whether each and every possible 
> configuration still works?

You cannot test each and every. But there are a couple of different
things that do happen:

I have just over 100 different configurations which i occassionally
compile. After a few hours of compiling i get a feeling that nothing
major is broken. I will sometimes do this for synth and pick a few
tests are random to actually run. One day i will add some scripts to
run them all...

eCosCentric have a test farm. This takes the concept a step
further. It builds and runs many more different configurations and
runs them on real hardware. If you are an eCosCentric customer, the
release you get has been thorough this test farm.

The choice is yours. You have the possibility to run the tests
yourself for your specific configuration on your hardware. Or you can
pay somebody like eCosCentric to give you a certified tested version.

> Perhaps an eCos release process needs a lot of community involvement. ie. 
> several release candidates made and all users requested to give it try in 
> the configuration they use. ie. Rather than attempting test coverage of 
> all possible configurations, at least get coverage of all existing usages.

It is not just testing. In the past a release generally also means a
new toolchain. This itself takes a lot of effort to build and
test. Then you need to build and test eCos with the new toolchain....
So an eCos release is a lot of work.

Having said that there is a new release planned, that will occur once
the transfer of eCos to the FSF is completed. But this is taking a lot
longer than we expected.
 
> I'm also sure there exist patches and bug fixes on peoples disks that the 
> announcement of a forthcoming release may shake loose.

Possibly, but after 3 years the regular users know that releases don't
happen very often and that they should use anoncvs. So they tend to
contribute there patches so they get included into anoncvs. 

        Andrew

-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss