Experimental Physics and
Industrial Control System

Andrew Johnson <[email protected]> · Wed, 30 Jun 2010 10:55:40 -0500

Hi Jeff,

On Monday 28 June 2010 19:51:17 Jeff Hill wrote:
>
> I admit to being a bit confused by this. As I recall, in the past the
> "Known Problems" patches were only quick fixes for a very limited set of
> issues, and these patches were intended for application only against the
> latest R3.14 branch.

Strictly speaking they were intended for application against the release 
version for which they were published (you imply the branch tip above, which 
would not be very useful).

> Presumably, under a scenario where R3.14 point release
> contain new features, the net result would contain also the new features in
> addition to all of the patches (some of the patches coming from the "Known
> Problems" patches and the rest being already included in the point
> release).

The Known Problems patches provided sites with fixes to problems discovered 
since the last release was made, but they did require that sites keep up to 
date with releases if they wanted to use them.  We don't have enough manpower 
to back-port fixes to old releases, but if a site needs a fix badly enough 
they could always do the back-porting themselves.

> The bottom line is that if we are to successfully (therefore efficiently)
> produce both a patched version and a new features version then presumably
> we would need to keep these in different branches in the source code
> control?

How far back would you want to keep those patched versions?  There would not 
be *a* patched version, we'd have to have 11 different branches, one for each 
released 3.14.x version, and would need to back-port all bug-fixes to every 
branch to which they apply.  I don't have the time to do that whenever I find 
and fix a bug, but as I said above EPICS sites can do the back-porting 
themselves if they want to.

They way that Linux distributions manage this combinatorial problem is they 
nominate certain releases for Long Term Support (LTS), and all other releases 
drop out of being maintained within a much shorter period.  If we had enough 
people available to do the work we could adopt that kind of strategy, but we 
don't.

- Andrew
-- 
The best FOSS code is written to be read by other humans -- Harald Welte

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System