[ESS] Current version?

Martin Maechler maechler at stat.math.ethz.ch
Tue Jan 24 22:23:17 CET 2012


On Tue, Jan 24, 2012 at 18:29, Rodney Sparapani <rsparapa at mcw.edu> wrote:
> Ken Williams wrote:
>>
>> Or perhaps I'm mistaken - does 5.14 actually change over time with
>> "stealth" updates that fix any bugs found?
>>
>
> Like I said, releases tend to be pretty stable.  But, when a new bug is
> identified, it is fixed first in the development branch; and you can get
> it there if you need it.  If the bug is fairly widespread, then a
> corresponding new release is triggered.
>
>> I'm not trying to criticize, I know it takes extra effort to maintain 2
>> branches of a codebase.  Just trying to explain why some non-core developer
>> might want to try 5.15 if they're looking for something vaguely like "the
>> thing that might work best for me right now".  And that if the dev branch is
>> not considered suitable for testing in situ, there probably won't be as much
>> testing of it.
>>
>
> We would appreciate it if people do test the development branch.  But,
> we want to stress that there may be issues; please be prepared to retreat
> to the released version.  This often means maintaining two separate
> setups.  However, the downside is as you suggest; testing by non-developers
> is infrequent.

I agree with Rodney... and also with most of what Ken said.
To get back to the issue of bug and bug reports, I just want to reiterate:

- *PLEASE* do report every bug you find in the released version of ESS.
  Those are the ones we (and yo!) really want to see fixed.

- The current 5.14 vs. "5.15 unstable/on the developement"  situation
has been very unusual.
  Indeed we currently have several largish new features in 5.15...
  Compared with the traditional work and release cycle of ESS, this is
new.  Typically we had added a few features, fixed a few bugs and then
made a new release.. and so the devel- version has never been far away
from the last release,... and hence bug reports on the released
version almost always would be fixed in the devel version, but the fix
would have been identical in the release...

I think we should think carefully before changing that cautious and
simple but also quite stable development process. We could start
working with a *patched and a *devel version, the same way as R is
developed.
But note that definitely needs a considerably larger group of both
developers and  "sophisticated-enough" users who are (by themselves!)
willing and loving to work with patched and or devel versions of
ESS....
At the moment I feel we are still not close to that, and the danger of
a large overhead for us (ESS core), of simultaneously maintaining
patch branch and a devel branch, of being disciplined in mergigng
things back from devel to patched of  regularly testing *both*
versions ... al that could really  destabilize developement rather
than stabilizing it....
The big drawback of ESS compared to R is that you cannot easily script
testing of ESS, in a similar way we (R-core) have been able to do that
for R and every R package...

Martin



More information about the ESS-help mailing list