maven-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Dennis Lundberg <denn...@apache.org>
Subject Re: [VOTE] Release Apache Maven Model Converter version 2.3
Date Fri, 16 Aug 2013 07:10:28 GMT
On Fri, Aug 16, 2013 at 1:20 AM, sebb <sebbaz@gmail.com> wrote:

> On 15 August 2013 20:57, Dennis Lundberg <dennisl@apache.org> wrote:
> > On Thu, Aug 15, 2013 at 9:27 PM, sebb <sebbaz@gmail.com> wrote:
> >
> >> On 15 August 2013 14:16, Jason van Zyl <jason@tesla.io> wrote:
> >> > What Sebb is doing is perfectly reasonable.
> >>
> >
> > I agree. Checking that the source bundle is correct is good release
> review
> > practice.
> >
> > Thank you!
> >>
> >> > He's trying to assert that everything in the source ball actually
> comes
> >> from source control and that no errant files have made their way into
> the
> >> distribution. Right now we cannot assert that the assembly plugin has
> not
> >> wandered outside the checkout and pulled something else in, or that
> someone
> >> didn't accidentally put something else in the distribution. I think this
> >> unlikely but we can't assert otherwise right now which I believe is
> Sebb's
> >> point.
> >>
> >> It *has* already happened several times that I am aware of.
> >>
> >> The last few releases of the War plugin (various RMs & voters)
> >> included at least one spurious file.
> >> So it was not just a one-off packaging - and review - failure.
> >> [See separate thread(s) for all the details; they are not germane here.]
> >>
> >> The message is that mistakes happen even in automated processes.
> >> Which is why independent comparison of input and output is valuable.
> >>
> >> > If we can embed the revision from which the assembly was made in the
> >> assembly itself (and maybe the build number plugin is doing this
> already),
> >> then a tool can be made to unpack the assembly, checkout the revision
> and
> >> assert that everything in the source distribution comes from source
> control.
> >> >
> >> > If we can also assert that as part of each build all the license files
> >> are intact and headers are in place then I believe we're done with
> >> provenance.
> >> > Licenses are present, all files have valid license headers, all files
> >> present in the source distribution come from source control, all
> >> contributions to source control are from bonafide CLA carrying Apache
> >> committers because you don't get access to commit until the CLA is on
> file.
> >> >
> >> > Sebb, reasonably accurate?
> >>
> >> Yes.
> >>
> >> One other point I made already is that I think the vote e-mail needs
> >> to be transparent to all, not just those on the PMC.
> >> Links to the output from the release process are obviously already in
> >> the mail; what is missing is the input to the process, e.g.. SCM
> >> coords.
> >> Yes, it may be possible to dig out the details from the archive, but
> >> that's not trivial.
> >>
> >
> > I disagree.
> >
> > If we focus first on a "normal" release, one that succeeds on the first
> > attempt, without a respin or deleting of tags.
> > To check provenance you would do this:
> >
> > 1. download the source bundle
> > 2. unpack the source bundle
> > 3. checkout the corresponding source code from the SCM
> > 4. compare the two trees
> >
> > Right so far?
> >
> > What you want, if I understand you correctly, is to have the SCM URL in
> the
> > vote email. So that you can give that to your SCM client in step 3.
>
> Yes.
>
> > With the process we use here at the Maven project, the SCM URL is in the
> > pom.xml file that sits in the root directory of the unpacked source
> bundle.
> > All you need to do is open the file and copy the URL from there. I still
> > fail to see how that is so much harder than to copy the URL from an
> email.
> >
> > That is if you don't know the conventions that we use, by way of the
> > Release Plugin. The tag will always be in the format
> > ${project.artifactId}-${project.version}
> >
>
> My point is that it should be completely transparent, even to outside
> reviewers.
>

I guess that this is the point that we'll have to agree to disagree on. My
view is that if someone wants to to review a release from the Maven
project, they'd have to have a basic understanding of how Maven works and
how we do releases in the Maven project. That includes what the Release
Plugin is and how it works.

Most people have their own checklist of stuff they do when reviewing a
release. Would it be a good idea if we aggregate all those points on a page
on the Maven web site, under the development section? That would also serve
as a guide for "outside reviewers".


> > Now, for a respun release thing are trickier. Here it might be a good
> idea
> > to include the revision number or hash, or whatever is unique in the SCM
> > being used.
>
> And how do you know from a vote e-mail that it is respun?
>

It should always say so in the subject of the vote email. Not sure if that
is written down somewhere, but that's how we have always done it. If it's
missing I'll add it to our release process document.


>
> > Even though the code under review will always be under the
> > latest tag in the above format (at least for Subversion).
>
> Until the next respin.
>
> If there is a respin, and reviewers are not following the e-mails very
> carefully, it would be quite easy to overlook an updated tag.
>
> This is all about making sure that it is really obvious what the vote is
> about.
>

Yes, that's why I agreed that it's a good idea to add the revision/hash
when doing a respin.


>
> >
> >> Publishing the SCM coordinates in the mail is trivial to do, and shows
> >> that the input is an important part of the review process.
> >> Having the information in the mail thread is also useful for the
> archives.
> >>
> >
> > As others have said before, we aim to automate the release process as
> much
> > as possible. Therefor even a seemingly minor addition will be questioned,
> > as you have noticed, before it is included in our process.
> >
> > Can you explain why the information is useful for the archives? I've seen
> > you mentioned that before. Isn't that moot once the release is approved?
> > The tag will be in Subversion for the forseable future and noone will
> touch
> > it. What am I missing?
>
> Why would a release need to be revisited?
> Perhaps someone is complaining that one of our releases contains code
> it should not.
> If that is the case, it helps to have the evidence of the release vote
> in plain sight.
>

Sure it helps in those cases. But the evidence in our case is in the the
source bundle itself, which is even better in my opinion. There you have
the tag in the POM file.
Note that I'm not talking about respins here, that's another story.


>
> >
> >>
> >> > On Aug 15, 2013, at 9:01 AM, Chris Graham <chrisgwarp@gmail.com>
> wrote:
> >> >
> >> >>
> >> >>
> >> >> Sent from my iPhone
> >> >>
> >> >> On 15/08/2013, at 10:05 PM, sebb <sebbaz@gmail.com> wrote:
> >> >>
> >> >>> On 15 August 2013 10:08, Chris Graham <chrisgwarp@gmail.com>
wrote:
> >> >>>> What sebb does not appear to have understood or accepted, as
> Stephen
> >> has
> >> >>>> endlessly pointed out, is that we vote on the source bundle,
not a
> scm
> >> >>>> revision, and that, strictly speaking a SCM is not even required
> >> (however
> >> >>>> sensible it is to use one).
> >> >>>>
> >> >>>> He wants a tree and a revision so that we can compare between
> >> releases,
> >> >>>> where what he should be doing, strictly speaking, is comparing
> source
> >> tar
> >> >>>> balls, as that is what we really are voting on.
> >> >>>
> >> >>> I agree that what is released (and voted on) are the source
> tarballs.
> >> >>> And any such tarballs should be identical (barring possibly
> different
> >> >>> EOL settings for text files).
> >> >>>
> >> >>> However, that is only one of the checks that need to be made.
> >> >>>
> >> >>> The PMC also needs to ensure that the files are being released
under
> >> >>> the correct license.
> >> >>
> >> >> Are not the licenses in the source that is in the source tarball?
> >> >>
> >> >> If so, can not the rat plugin or similar be used to check the
> >> compliance?
> >> >>
> >> >>> I contend that the only practical way to check the licences is
to
> >> >>> compare the source tarball(s) with the files in SCM.
> >> >>>
> >> >>> [The files should only be added to SCM if the license is OK, so
the
> >> >>> SCM tag acts as a database of validated source files.]
> >> >>>
> >> >>> The SVN revision / Git hash are needed to ensure uniqueness.
> >> >>>
> >> >>>
> ---------------------------------------------------------------------
> >> >>> To unsubscribe, e-mail: dev-unsubscribe@maven.apache.org
> >> >>> For additional commands, e-mail: dev-help@maven.apache.org
> >> >>>
> >> >>
> >> >> ---------------------------------------------------------------------
> >> >> To unsubscribe, e-mail: dev-unsubscribe@maven.apache.org
> >> >> For additional commands, e-mail: dev-help@maven.apache.org
> >> >>
> >> >
> >> > Thanks,
> >> >
> >> > Jason
> >> >
> >> > ----------------------------------------------------------
> >> > Jason van Zyl
> >> > Founder,  Apache Maven
> >> > http://twitter.com/jvanzyl
> >> > ---------------------------------------------------------
> >> >
> >> > There's no sense in being precise when you don't even know what you're
> >> talking about.
> >> >
> >> >  -- John von Neumann
> >> >
> >> >
> >> >
> >> >
> >> >
> >> >
> >>
> >> ---------------------------------------------------------------------
> >> To unsubscribe, e-mail: dev-unsubscribe@maven.apache.org
> >> For additional commands, e-mail: dev-help@maven.apache.org
> >>
> >> --
> >> Dennis Lundberg <dev-help@maven.apache.org>
> >>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@maven.apache.org
> For additional commands, e-mail: dev-help@maven.apache.org
>
> --
> Dennis Lundberg <dev-help@maven.apache.org>
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message