Folks -
Well, I pretty much stay out of the fray on issues of amber code
development, and I don't know that I should voice an opinion, given that I
have pointedly keep pmemd out of the fray and really try to not be involved
in anything where careless checkins will make my life miserable. However, I
would like to voice strong support for English Bob in this little
discussion. By and large, the amber development environment has
uncontrolled interfaces, responsibilities, etc. etc. In my experience in
industry, this is a really bad idea, and you don't often encounter
situations where a half a dozen guys are hacking simultaneously on the same
source files. It's untenable, and the organizational cost of people
stepping on each other is just huge. So given our situation, I think that
"be careful, or else..." is actually a little mild. We really do need, at
some point, to do a better job of factoring the code that is gang-developed,
and defining the interfaces, as it were. I can't afford the time to get
anywhere near it, and besides, being the cop on anything really brings out
the worst in me, and I hate it. So all I can say at this point is that Ross
has every right to complain, that people developing in this environment need
to be very very very careful, and that when you do something that blows away
everybody else, you basically waste a heck of a lot more time that belongs
to other people than you wasted of your own time. At microsoft they
actually made folks who broke the build wear a dunce cap for the day, hoping
the social pressure would get them to clean up their act. By the time I was
there, they had gotten a bit more politically correct, and had reverted to
just having half your management chain and your coworkers come into your
office and take your head off; I expect it affected bonuses also. We should
all take this stuff very seriously.
Regards, and with appreciation to Ross for sweeping up the mess - Bob
----- Original Message -----
From: "Scott Brozell" <sbrozell.scripps.edu>
To: <amber-developers.scripps.edu>
Sent: Thursday, October 12, 2006 10:43 PM
Subject: Re: amber-developers: Current CVS Tests Status
> Hi,
>
> On Thu, 12 Oct 2006, Ross Walker wrote:
>
>> Hi All,
>>
>> we should then setup a machine that checks out the CVS tree everynight
>> and runs the full test suite both in serial and parallel and reports any
>> problems. I have a machine that we can use for this.
>
> There is already a script for nightly testing: test/nightly_test
> It was used regularly with amber8, and Ive been meaning to restart it
> on pusar ( ./configure sgi_mips )
> Since more testing is better testing, I recommend that you use it on
> your machine too.
>
>> Then if the next morning we find that somebodies changes broke things in
>> either serial or parallel we can unwind those changes.
>
> Unwinding is probably too unilateral; a friendly email to the committer(s)
> has worked in the past.
>
>> Can I also 'politely' ask that before ANYBODY checks anything into the
>> cvs
>> tree they test everything thoroughly. This means you do a FULL build and
>> run
>> all test cases both in SERIAL AND IN PARALLEL... somebody's recent
>> changes
>
> This should be standard operating procedure. (But we humans arent
> perfect.)
> And while the soapbox is available:
> please both cvs update frequently and commit your tested development code
> frequently. This substantially improves the quality control of Amber.
>
>> Followed by castration with a rusty spoon without anaesthetic ;-)
>
> Ouch, there's a new sheriff in town, he's mean, and his name is English
> Bob !
> err Ross :)
>
> Scott
>
>
Received on Sun Oct 15 2006 - 06:07:10 PDT
