commons-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Gilles Sadowski <gil...@harfang.homelinux.org>
Subject Re: [VOTE][RC5] Release Commons Math 3.1
Date Fri, 21 Dec 2012 19:16:46 GMT
On Fri, Dec 21, 2012 at 01:13:48PM +0100, Gilles Sadowski wrote:
> On Fri, Dec 21, 2012 at 06:40:36AM -0500, Gary Gregory wrote:
> > On Dec 21, 2012, at 4:18, Thomas Neidhart <thomas.neidhart@gmail.com> wrote:
> > 
> > > On Thu, Dec 20, 2012 at 11:26 AM, Gilles Sadowski <
> > > gilles@harfang.homelinux.org> wrote:
> > >
> > >> Hi.
> > >>
> > >> Please have a look at the next candidate (RC5), and vote for the release
> > >> of Commons Math 3.1.
> > >>
> > >> ----------
> > >> Tag:
> > >>  https://svn.apache.org/repos/asf/commons/proper/math/tags/MATH_3_1_RC5/
> > >>
> > >> Site:
> > >>  http://people.apache.org/builds/commons/math/3.1/RC5/
> > >>
> > >> Binaries:
> > >>
> > >> https://repository.apache.org/content/repositories/orgapachecommons-052/org/apache/commons/commons-math3/3.1/
> > >>
> > >> [ ] +1 Release it.
> > >> [ ] +0 Go ahead; I don't care.
> > >> [ ] -0 There are a few minor glitches: ...
> > >> [ ] -1 No, do not release it because ...
> > >>
> > >> This vote will close in 72 hours.
> > >> ----------
> > >
> > > Hi,
> > >
> > > this is not yet my vote, just a request for clarification.
> > >
> > > I checked again the Clirr errors, and there are still the two related to
> > > the probability method in LogNormal and NormalDistribution.
> > >
> > > In 3.0, there was a method with only 1 parameter, which always returned 0.
> > > Now there is a probability method with two parameters, which is defined in
> > > the implemented interface AbstractRealDistribution.
> > >
> > > You mentioned that this is a false positive, but I doubt this. Maybe the
> > > probability method was never used, but then it should at least be mentioned
> > > in the release notes.
> > >
> > > Sorry to be pedantic about this.
> > 
> > Clirr lists 7 errors, so strictly speaking it does not looks like 3.1
> > is binary compatible with 3.0. I see these options:
> 
> At the time of deleting the "enum" fields ("ALPHA", "BETA", etc), we got a
> green light, on the basis that those fields were only meant for CM's
> internal use.
> 
> You are certainly right that someone who actually use those in his
> application will get into trouble; but then we knew that when we discussed
> the removal. We agreed that this situation was clear and not to be taken
> into account.
> 
> For the reports on the method signature change, I've answered to Thomas's
> post.
> Could please someone point to a source other than Clirr if it is not
> satisfactory?
> 
> > 
> > 1) document 3.1 as not BC. This is not what we usually do in Commons.
> > 2) fix the clirr errors in the code. This is the safe option.
> > 3) make this a 4.0 releases and change the packages and maven
> > coordinates. This seems like a high price and likely not what the
> > [math] community intends for a real 4.0.
> > 

So? In the unlikely event that someone complains about missing fields from
the "LocalizedFormats" enum, do we assume that binary compatibility does not
cover such things as using CM's "internal" classes (even if we cannot
enforce this restriction in Java)?


Gilles

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
For additional commands, e-mail: dev-help@commons.apache.org


Mime
View raw message