[MITgcm-support] doc/tag-index

[Chris Hill]

Hi Holly and others,

 Since Martin thinks someone is more on top of things than him (which
I'm not sure is true) I'll say something on
 what we try and do.

 We found that a so-called stable release wasn't that interesting to
most people (even though it sounds
 good on paper and to the managerial oriented). Mostly what people wanted was
 a stable release and the newest features! So we have a system of
continuous updates
 with a labeling system checkpointNN[a-z]. Generally something that is labelled
 "checkpointNN[a-z]" is doing reasonably well and isn't total junk.
Increments in the [a-z] letter appear
 on weeks to  months time scales. Increments in the NN suffix appear
less frequently, months apart
 generally. A checkpointNN thing is usually better tested (especially
with regard to
 AD features) and usually marks a boundary between more invasive bits
of development.

 Most people take a recent "checkpoinNN[a-z]" thing and then stick
with that for a while.
 That helps in knowing what you have, being able to reproduce things
and in asking questions, since
 its clear what you started with. Once a thing is labelled
checkpointNN[a-z] it is frozen and a
 checkout of checkpointNN[a-z] will always give the same thing.
 Other people take the most recent thing and continually update, but
thats for the
 more adventurous and fearless folks.

Chris

On Wed, Aug 5, 2009 at 10:53 AM, Martin Losch<Martin.Losch at awi.de> wrote:
> Holly,
>
> you don't seem to get an answer from those whom I consider more on top of
> this issue than I am. I hope that I am not giving away any secrets about the
> MITgcm ...
>
> doc/tag-index is a reference file where the developers briefly document
> their changes in the code; so you are right that it is actually quite useful
> for finding out about the status of the MITgcm.
>
> There is no "stable" version (there was a Release1 years ago, but never a
> second one).
> The latest code is tested every day and there are a few (unnamed)
> bloodhounds who will immediately notice any problems with these tests, so
> that the latest code is almost always OK for the parts that are used in the
> verification experiments (the developers have the strict rule to only check
> in modifications after running these tests).
> The normal user should be fine with the latest code. In fact if you use it
> you make sure that you have the latest bug-fixes etc include (but it also
> means that you get new code which might contain new bugs). However, before
> you update a running project to latest code I advise you to check in
> tag-index that this update will not change your results (e.g. there have
> been major changes in the seaice package around checkpoint61o/p/q/r, which
> do change the results if you use pkg/seaice, but these events are rare).
>
> I guess it is usually safe to use "checkpoints", such as checkpoint61.
> Again: we do our best (see above) to have properly working code at all times
> and particularly at the time of "tags" (e.g. currently the latest tag is
> checkpoint61t).
>
> Martin
>
> On Aug 3, 2009, at 11:25 PM, Holly Dail wrote:
>
>> Hello all -
>>
>> When I do a fresh checkout from the CVS, I'm never sure I'm getting a
>> reasonable version.  After recent mentions of doc/tag-index I looked at it
>> for the first time -- it looks like a useful way to find reasonably stable
>> versions to focus on.  Is checking out the most recent checkpoint mentioned
>> in that file my best bet?  Do all developers update it or is it sporadic?
>>
>> Thanks -
>> Holly
>>
>> PS I was going to ask how to search the archives, because its been
>> plaguing me.  And then I thought of trying Google:
>> site:http://forge.csail.mit.edu/pipermail/mitgcm-support my question
>> Thought I'd mention it in case anyone else misses the obvious too.
>> _______________________________________________
>> MITgcm-support mailing list
>> MITgcm-support at mitgcm.org
>> http://mitgcm.org/mailman/listinfo/mitgcm-support
>
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-support
>