Re: [RFA] deleting breakpoints inside of 'commands' [Repost]

[Kevin Buettner <kevinb at cygnus dot com>]

On Sep 26,  4:01pm, Fernando Nasser wrote:

> And I don't think Don's latest patch is complex, or considerably more
> complex than the simplistic copy approach (or hack!).  It is elegant 
> (although Kevin's comments are valid and should be incorporated).
> 
> If Don can add a cleanup function and do the polishing suggested by Kevin
> on his last patch I suggest that we stick with that one.

Here are my preferences...

I like the patch that makes a copy the best since the changes are easy
to understand and localized to one function.

Next best would be a patch (unwritten as of now) which implements
reference counts for these command chains.  I think that a lot of
folks understand reference counts and it takes less time to grok code
written with this design pattern.  In all likelyhood, such a patch
would look somewhat similar to Don's latest patch though; i.e, the
changes would be distributed among several files and would likely be
in the same functions that Don has touched with his latest patch.

Next best would be Don's latest patch revised to use an enum for
the ``execute'' member.  I think this code would be easier to
understand by giving the 0, 1, and 2+ values actual names that are
meaningful.  I still have concerns about what might happen if
bpstat_do_actions() is ever executed recursively...  I don't think this
ever happens, so it's probably a non-issue, but if it could happen,
then a reference counting mechanism would definitely be superior.

(All of the above approaches need to make use of a cleanup.)

Kevin

P.S. I'd like to publicly thank Don for all the hard work he has
put in on this patch.  I imagine it must be frustrating to be told
to do the patch one way by one person, a different way by a second,
and still a different way by a third...