You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@maven.apache.org by Karl Heinz Marbaise <kh...@gmx.de> on 2016/06/12 11:38:04 UTC
https://issues.apache.org/jira/browse/MNG-5227
Hi,
based on the feedback of the tests for current Maven 3.4.0-SNAPSHOT
testing I have taken a deeper look into some of the changes.
Based on the above issue we now have changed the behaviour of Maven
3.4.0 compared to 3.3.9 (and before).
It looks like this change breaks some builds...
So the question is:
* Is this the correct path we are going?
Handling test scope in a more strict way than before?
* I think we need to improve some of the docs to make it more clear
and of course add this into the release notes with
a red light in there...
WDYT ?
Kind regards
Karl Heinz
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@maven.apache.org
For additional commands, e-mail: dev-help@maven.apache.org
Re: https://issues.apache.org/jira/browse/MNG-5227
Posted by Christian Schulte <cs...@schulte.it>.
Am 06/12/16 um 15:40 schrieb Robert Scholte:
>>
>> Why?
>>
>
> Let me quote myself:
> "I've been talking about this change with a couple of fellow committers at
> FOSDEM and it convinced me that you never ever want to manage optional.
> They have too much effect on the dependencies using it. For dependencies
> which are expected to be non-optional (default) it suddenly will be.
If you say so, yes. That's the point. It's you managing that flag. So
that's not a "suddenly" that's an "intentional because you say so".
There are quite some use cases you would want to manage the optional
flag. Only drawback is projects having used the optional flag in
dependency management not noticing that this is ignored before 3.4.
That's the only issue I am seeing with this.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@maven.apache.org
For additional commands, e-mail: dev-help@maven.apache.org
Re: https://issues.apache.org/jira/browse/MNG-5227
Posted by Robert Scholte <rf...@apache.org>.
On Sun, 12 Jun 2016 15:10:35 +0200, Christian Schulte <cs...@schulte.it>
wrote:
> Am 06/12/16 um 15:01 schrieb Robert Scholte:
>> Based on the description "Make 'optional' flag of a dependency
>> manageable"
>> I would say -1 for this change.
>
> Why?
>
Let me quote myself:
"I've been talking about this change with a couple of fellow committers at
FOSDEM and it convinced me that you never ever want to manage optional.
They have too much effect on the dependencies using it. For dependencies
which are expected to be non-optional (default) it suddenly will be.
Instead the ModelValidator should inform the users that setting the
optional-flag on a managed dependency has no effect."
>> Or is this the answer to global excludes?
>
> It can be used to exclude a dependency globally by flagging it optional
> in dependency management.
>
>> Maybe the description is misleading, because IIRC this is more about
>> transitive dependencies which were marked as optional. Those
>> dependencies
>> should not end up in the resolution.
>
> No. It's about allowing to manage the optional flag the same way the
> scope and version can be managed. What's wrong with that? It's a missing
> feature.
>
>> Give Christian Schulte the chance to clarify this again, based on 2 or 3
>> poms. Hopefully that explains it a bit more. Issue has reached TLDR
>> status.
>
> What is there I need to clarify?
>
>
> ---------------------------------------------------------------------
> 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
Re: https://issues.apache.org/jira/browse/MNG-5227
Posted by Karl Heinz Marbaise <kh...@gmx.de>.
Hi,
On 6/12/16 3:10 PM, Christian Schulte wrote:
> Am 06/12/16 um 15:01 schrieb Robert Scholte:
>> Based on the description "Make 'optional' flag of a dependency manageable"
>> I would say -1 for this change.
>
> Why?
At the moment (I'm not sure about this) is that this change is changing
behaviour in relationship to test scope dependencies which breaks builds
in comparsion to Maven 3.3.X and before...
That's why Robert says -1 ...
We need to clarify if this is related to this change or not...or do we
have other changes which can cause this...so we need to identify this...
>
>> Or is this the answer to global excludes?
>
> It can be used to exclude a dependency globally by flagging it optional
> in dependency management.
>
>> Maybe the description is misleading, because IIRC this is more about
>> transitive dependencies which were marked as optional. Those dependencies
>> should not end up in the resolution.
>
> No. It's about allowing to manage the optional flag the same way the
> scope and version can be managed. What's wrong with that? It's a missing
> feature.
Yes managing the optional flag is one thing but based on the code
examples you have given in the issue it looks like it influences the
test scope dependencies as well cause it's mentioned with "provided"
scope...
>
>> Give Christian Schulte the chance to clarify this again, based on 2 or 3
>> poms. Hopefully that explains it a bit more. Issue has reached TLDR status.
>
> What is there I need to clarify?
>
We need to check if this issue is causing those problems? (See users
list)...
After checking git history: Based on the code changes which are related
to the issue MNG-5227 it is only related to managing optional flag. And
not about test transitivity...So this means there is an other change
which caused the change in behaviour...
Kind regards
Karl Heinz Marbaise
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@maven.apache.org
For additional commands, e-mail: dev-help@maven.apache.org
Re: https://issues.apache.org/jira/browse/MNG-5227
Posted by Christian Schulte <cs...@schulte.it>.
Am 06/12/16 um 15:01 schrieb Robert Scholte:
> Based on the description "Make 'optional' flag of a dependency manageable"
> I would say -1 for this change.
Why?
> Or is this the answer to global excludes?
It can be used to exclude a dependency globally by flagging it optional
in dependency management.
> Maybe the description is misleading, because IIRC this is more about
> transitive dependencies which were marked as optional. Those dependencies
> should not end up in the resolution.
No. It's about allowing to manage the optional flag the same way the
scope and version can be managed. What's wrong with that? It's a missing
feature.
> Give Christian Schulte the chance to clarify this again, based on 2 or 3
> poms. Hopefully that explains it a bit more. Issue has reached TLDR status.
What is there I need to clarify?
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@maven.apache.org
For additional commands, e-mail: dev-help@maven.apache.org
Re: https://issues.apache.org/jira/browse/MNG-5227
Posted by Karl Heinz Marbaise <kh...@gmx.de>.
Hi Christian,
On 6/12/16 3:48 PM, Christian Schulte wrote:
> That will all go into the release notes, of course. There is more to
> come. Before we go any further. I've just pushed a MNG-6006 branch. Do this:
Which couldn't be done yet cause the sub tasks are not completed. As i
mentioned in the other threads things are not yet done. That's the state
of the thing.
Yes there are bugs in Aether but at the moment we can't push the new
Aether out cause things are not done yet. This could produce license
etc. violations etc. which i don't like to discuss with Herv� and the
Eclipse Board.
Unfortunately I need to say -1 for that as well...
Karl Heinz Marbaise
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@maven.apache.org
For additional commands, e-mail: dev-help@maven.apache.org
Re: https://issues.apache.org/jira/browse/MNG-5227
Posted by Christian Schulte <cs...@schulte.it>.
Am 06/12/16 um 15:01 schrieb Robert Scholte:
>> Hi,
>>
>> based on the feedback of the tests for current Maven 3.4.0-SNAPSHOT
>> testing I have taken a deeper look into some of the changes.
>>
>> Based on the above issue we now have changed the behaviour of Maven
>> 3.4.0 compared to 3.3.9 (and before).
>>
>> It looks like this change breaks some builds...
The changes are Aether bugfixes. That's not related to MNG-5227 in any
way. The builds rely on misbehaviour and need to be fixed.
>>
>> So the question is:
>>
>> * Is this the correct path we are going?
Yes. Fixing bugs is normal progress.
>>
>> Handling test scope in a more strict way than before?
It's not a more strict way. It's an Aether bugfix making Maven behave as
advertised.
>>
>>
>> * I think we need to improve some of the docs to make it more clear
>> and of course add this into the release notes with
>> a red light in there...
>>
>>
>> WDYT ?
That will all go into the release notes, of course. There is more to
come. Before we go any further. I've just pushed a MNG-6006 branch. Do this:
Build a current Aether version from here:
git clone https://git-wip-us.apache.org/repos/asf?p=maven-aether.git
mvn install
Then checkout the MNG-6006 branch and build a 3.4.0-SNAPSHOT with that
Aether version.
Use that 3.4.0-SNAPSHOT. We've allready progressed in fixing Aether bugs
months ago. There currently is no way to make those bugfixes appear
somewhere. With those bugfixes in place, you will discover various
things to fail due to bugs in the poms of those failing projects. For
example: Herv Boutemy asked me to provide an IT for
<https://issues.apache.org/jira/browse/MPLUGIN-296>. The answer to this
is: Use a recent Maven version with those Aether bugfixes in place and
see all existing ITs starting to fail. And it is correct those ITs are
failing with the Aether bugfixes in place.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@maven.apache.org
For additional commands, e-mail: dev-help@maven.apache.org
Re: https://issues.apache.org/jira/browse/MNG-5227
Posted by Robert Scholte <rf...@apache.org>.
Based on the description "Make 'optional' flag of a dependency manageable"
I would say -1 for this change. Or is this the answer to global excludes?
Even then I'm -1 and prefer extra elements in the model 5.
Maybe the description is misleading, because IIRC this is more about
transitive dependencies which were marked as optional. Those dependencies
should not end up in the resolution.
Give Christian Schulte the chance to clarify this again, based on 2 or 3
poms. Hopefully that explains it a bit more. Issue has reached TLDR status.
Robert
On Sun, 12 Jun 2016 13:38:04 +0200, Karl Heinz Marbaise
<kh...@gmx.de> wrote:
> Hi,
>
> based on the feedback of the tests for current Maven 3.4.0-SNAPSHOT
> testing I have taken a deeper look into some of the changes.
>
> Based on the above issue we now have changed the behaviour of Maven
> 3.4.0 compared to 3.3.9 (and before).
>
> It looks like this change breaks some builds...
>
> So the question is:
>
> * Is this the correct path we are going?
>
> Handling test scope in a more strict way than before?
>
>
> * I think we need to improve some of the docs to make it more clear
> and of course add this into the release notes with
> a red light in there...
>
>
> WDYT ?
>
> Kind regards
> Karl Heinz
>
> ---------------------------------------------------------------------
> 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