Re: Releasing Maven Resolver 1.1.0
I removed the test-util module: yes, not expected to be a runtime dependency. And this offers the opportunity to put the jar plugin configuration in a "java9-module" profile but I kept module name in javadoc group: we're not building with JDK 9 (and it currently does not work, requires some tuning), need a solution waiting for real Java 9 modules with JDK 9 javadoc output Regards, Hervé Le mardi 30 mai 2017, 20:16:59 CEST Robert Scholte a écrit : > I'm having my doubts if we should add the module name in the title of > Javadoc. > > Have a look at the Java 9 javadoc, which has a lot of module info "out of > the box". > > Robert > > On Tue, 30 May 2017 07:58:54 +0200, Robert Scholte > > wrote: > > How about ignoring the testutil? It should never become a runtime > > dependency, so I don't expect any project will refer to it, hence why > > give it a module name? > > > > Robert > > > > On Tue, 30 May 2017 01:26:52 +0200, Hervé BOUTEMY > > > > wrote: > >> commit proposed in a new branch > >> and generated site from this branch published for review: > >> - aggregated javadoc > >> https://maven.apache.org/resolver-archives/resolver-LATEST/apidocs/index. > >> html > >> > >> - API javadoc > >> https://maven.apache.org/resolver-archives/resolver-LATEST/maven-resolver > >> -api/ apidocs/index.html > >> > >> - Implementation javadoc > >> http://maven.apache.org/resolver-archives/resolver-LATEST/maven-resolver-> > >> >> impl/ apidocs/index.html > >> > >> Regards, > >> > >> Hervé > >> > >> Le mardi 30 mai 2017, 01:05:24 CEST Hervé BOUTEMY a écrit : > >>> another complementary reaction while reviewing consistency between java > >>> package names and modules names: perhaps we should change > >>> org.apache.maven.resolver.testutil > >>> to org.apache.maven.resolver.internal.test.util > >>> > >>> Regards, > >>> > >>> Hervé > >>> > >>> Le mardi 30 mai 2017, 00:50:51 CEST Hervé BOUTEMY a écrit : > >>> > one associated question I had in mind: how do we document to end > >>> > >>> users > >>> > >>> > what > >>> > are the module names? Should we add a report to MPIR? And how could > >>> > >>> this > >>> > >>> > report work, particularly on Automatic Module Name? > >>> > > >>> > I'm torn on choosing module name for this component: I think I > >>> > >>> understand > >>> > >>> > Stephen's logic. > >>> > Whatever name we choose, there will be a hard step when we move out > >>> > >>> of > >>> > >>> > org.eclipse.aether package in Maven Resolver 2.0. > >>> > > >>> > I'm not sure choosing org.eclipse.aether.* module names for Maven > >>> > >>> Resolver > >>> > >>> > 1.x and org.apache.maven.resolver.* module names for Maven Resolver > >>> > >>> 2.x > >>> > >>> > will ease the transition: whatever the choice, the tricky history of > >>> > >>> this > >>> > >>> > component will make its references tricky. > >>> > > >>> > While we do the trick at Maven artifact coords level, being > >>> > >>> consistent on > >>> > >>> > the same trick at module names. > >>> > One idea: perhaps adding the module names in consolidated javadoc > >>> > >>> groups > >>> > >>> > [1] may be a solution to document the hard choice. > >>> > > >>> > Regards, > >>> > > >>> > Hervé > >>> > > >>> > [1] > >>> > >>> http://maven.apache.org/resolver-archives/resolver-LATEST/apidocs/ > >>> > >>> > index.html > >>> > > >>> > Le lundi 29 mai 2017, 21:42:11 CEST Stephen Colebourne a écrit : > >>> > > Well, you have my opinion. I don't think there is an exemption here > >>> > > just because the component has a tricky history, and I personally > >>> > > think that any exemption for the package name necessarily applies > >>> > >>> to > >>> > >>> > > the module name (since it is now generally agreed that the module > >>> > >>> name > >>> > >>> > > derives from the package name). > >>> > > > >>> > > Doing otherwise will end up being confusing as more and more people > >>> > > name modules after super-packages. > >>> > > > >>> > > Stephen > >>> > > > >>> > > On 29 May 2017 at 18:46, Robert Scholte > >>> > >>> wrote: > >>> > > > This makes it an interesting case :) > >>> > > > > >>> > > > In short: the name "Aether" is owned by Eclipse and we are not > >>> > >>> allowed > >>> > >>> > > > to > >>> > > > use it. > >>> > > > However, we are allowed to use these packages for compatibility > >>> > > > reasons > >>> > > > as > >>> > > > long as needed. > >>> > > > > >>> > > > Module names are not part of this compatibility requirement, so > >>> > >>> we > >>> > >>> > > > shouldn't use the name Aether here. > >>> > > > Will there be an org.apache.maven.resolver package-based > >>> > > > implementation? > >>> > > > Not sure, but could very well be. > >>> > > > > >>> > > > Based on that I think we're now using the correct/preferred > >>> > >>> module > >>> > >>> > > > names. > >>> > > > > >>> > > > thanks, > >>> > > > Robert > >>> > > > > >>> > > > > >>> > > > On Mon, 29 May 2017 18:55:53 +0200, Stephen Colebourne > >>>
Re: Releasing Maven Resolver 1.1.0
I'm having my doubts if we should add the module name in the title of Javadoc. Have a look at the Java 9 javadoc, which has a lot of module info "out of the box". Robert On Tue, 30 May 2017 07:58:54 +0200, Robert Scholte wrote: How about ignoring the testutil? It should never become a runtime dependency, so I don't expect any project will refer to it, hence why give it a module name? Robert On Tue, 30 May 2017 01:26:52 +0200, Hervé BOUTEMY wrote: commit proposed in a new branch and generated site from this branch published for review: - aggregated javadoc https://maven.apache.org/resolver-archives/resolver-LATEST/apidocs/index.html - API javadoc https://maven.apache.org/resolver-archives/resolver-LATEST/maven-resolver-api/ apidocs/index.html - Implementation javadoc http://maven.apache.org/resolver-archives/resolver-LATEST/maven-resolver-impl/ apidocs/index.html Regards, Hervé Le mardi 30 mai 2017, 01:05:24 CEST Hervé BOUTEMY a écrit : another complementary reaction while reviewing consistency between java package names and modules names: perhaps we should change org.apache.maven.resolver.testutil to org.apache.maven.resolver.internal.test.util Regards, Hervé Le mardi 30 mai 2017, 00:50:51 CEST Hervé BOUTEMY a écrit : > one associated question I had in mind: how do we document to end users > what > are the module names? Should we add a report to MPIR? And how could this > report work, particularly on Automatic Module Name? > > I'm torn on choosing module name for this component: I think I understand > Stephen's logic. > Whatever name we choose, there will be a hard step when we move out of > org.eclipse.aether package in Maven Resolver 2.0. > > I'm not sure choosing org.eclipse.aether.* module names for Maven Resolver > 1.x and org.apache.maven.resolver.* module names for Maven Resolver 2.x > will ease the transition: whatever the choice, the tricky history of this > component will make its references tricky. > > While we do the trick at Maven artifact coords level, being consistent on > the same trick at module names. > One idea: perhaps adding the module names in consolidated javadoc groups > [1] may be a solution to document the hard choice. > > Regards, > > Hervé > > [1] http://maven.apache.org/resolver-archives/resolver-LATEST/apidocs/ > index.html > > Le lundi 29 mai 2017, 21:42:11 CEST Stephen Colebourne a écrit : > > Well, you have my opinion. I don't think there is an exemption here > > just because the component has a tricky history, and I personally > > think that any exemption for the package name necessarily applies to > > the module name (since it is now generally agreed that the module name > > derives from the package name). > > > > Doing otherwise will end up being confusing as more and more people > > name modules after super-packages. > > > > Stephen > > > > On 29 May 2017 at 18:46, Robert Scholte wrote: > > > This makes it an interesting case :) > > > > > > In short: the name "Aether" is owned by Eclipse and we are not allowed > > > to > > > use it. > > > However, we are allowed to use these packages for compatibility > > > reasons > > > as > > > long as needed. > > > > > > Module names are not part of this compatibility requirement, so we > > > shouldn't use the name Aether here. > > > Will there be an org.apache.maven.resolver package-based > > > implementation? > > > Not sure, but could very well be. > > > > > > Based on that I think we're now using the correct/preferred module > > > names. > > > > > > thanks, > > > Robert > > > > > > > > > On Mon, 29 May 2017 18:55:53 +0200, Stephen Colebourne > > > > > > wrote: > > >> The module name should in almost all cases be the super-package of > > >> the > > >> project. > > >> Don't use underscores in the module name unless they are also used in > > >> the package name. > > >> > > >> If the super-package is "org.apache.maven.resolver.api" then that is > > >> what the module name should be. > > >> But if the super-package is "org.apache.maven.resolver" then that is > > >> what the module name should be. > > >> > > >> ie. don't add ".api" just because that is what the artifact is > > >> called. > > >> Artifact name != module name. > > >> > > >> However, when I look at the source tree, it seems that the > > >> super-package name is "org.eclipse.aether". If I'm right, then that > > >> should be the module name. And maven-resolver-impl is a problem > > >> because it has two super-packages, but the module name should > > >> probably > > >> be "org.eclipse.aether.impl" with the internal package moved under > > >> impl. > > >> > > >> In summary, ignore the artifact name! Its the package name that > > >> matters when defining the module name. > > >> > > >> Stephen > > >> > > >> On 27 May 2017 at 17:43, Robert Scholte wrote: > > >>> There's no experience with this yet. > > >>> > > >>> Stephen Colebourne has written to related blogs: module naming
Re: Releasing Maven Resolver 1.1.0
How about ignoring the testutil? It should never become a runtime dependency, so I don't expect any project will refer to it, hence why give it a module name? Robert On Tue, 30 May 2017 01:26:52 +0200, Hervé BOUTEMY wrote: commit proposed in a new branch and generated site from this branch published for review: - aggregated javadoc https://maven.apache.org/resolver-archives/resolver-LATEST/apidocs/index.html - API javadoc https://maven.apache.org/resolver-archives/resolver-LATEST/maven-resolver-api/ apidocs/index.html - Implementation javadoc http://maven.apache.org/resolver-archives/resolver-LATEST/maven-resolver-impl/ apidocs/index.html Regards, Hervé Le mardi 30 mai 2017, 01:05:24 CEST Hervé BOUTEMY a écrit : another complementary reaction while reviewing consistency between java package names and modules names: perhaps we should change org.apache.maven.resolver.testutil to org.apache.maven.resolver.internal.test.util Regards, Hervé Le mardi 30 mai 2017, 00:50:51 CEST Hervé BOUTEMY a écrit : > one associated question I had in mind: how do we document to end users > what > are the module names? Should we add a report to MPIR? And how could this > report work, particularly on Automatic Module Name? > > I'm torn on choosing module name for this component: I think I understand > Stephen's logic. > Whatever name we choose, there will be a hard step when we move out of > org.eclipse.aether package in Maven Resolver 2.0. > > I'm not sure choosing org.eclipse.aether.* module names for Maven Resolver > 1.x and org.apache.maven.resolver.* module names for Maven Resolver 2.x > will ease the transition: whatever the choice, the tricky history of this > component will make its references tricky. > > While we do the trick at Maven artifact coords level, being consistent on > the same trick at module names. > One idea: perhaps adding the module names in consolidated javadoc groups > [1] may be a solution to document the hard choice. > > Regards, > > Hervé > > [1] http://maven.apache.org/resolver-archives/resolver-LATEST/apidocs/ > index.html > > Le lundi 29 mai 2017, 21:42:11 CEST Stephen Colebourne a écrit : > > Well, you have my opinion. I don't think there is an exemption here > > just because the component has a tricky history, and I personally > > think that any exemption for the package name necessarily applies to > > the module name (since it is now generally agreed that the module name > > derives from the package name). > > > > Doing otherwise will end up being confusing as more and more people > > name modules after super-packages. > > > > Stephen > > > > On 29 May 2017 at 18:46, Robert Scholte wrote: > > > This makes it an interesting case :) > > > > > > In short: the name "Aether" is owned by Eclipse and we are not allowed > > > to > > > use it. > > > However, we are allowed to use these packages for compatibility > > > reasons > > > as > > > long as needed. > > > > > > Module names are not part of this compatibility requirement, so we > > > shouldn't use the name Aether here. > > > Will there be an org.apache.maven.resolver package-based > > > implementation? > > > Not sure, but could very well be. > > > > > > Based on that I think we're now using the correct/preferred module > > > names. > > > > > > thanks, > > > Robert > > > > > > > > > On Mon, 29 May 2017 18:55:53 +0200, Stephen Colebourne > > > > > > wrote: > > >> The module name should in almost all cases be the super-package of > > >> the > > >> project. > > >> Don't use underscores in the module name unless they are also used in > > >> the package name. > > >> > > >> If the super-package is "org.apache.maven.resolver.api" then that is > > >> what the module name should be. > > >> But if the super-package is "org.apache.maven.resolver" then that is > > >> what the module name should be. > > >> > > >> ie. don't add ".api" just because that is what the artifact is > > >> called. > > >> Artifact name != module name. > > >> > > >> However, when I look at the source tree, it seems that the > > >> super-package name is "org.eclipse.aether". If I'm right, then that > > >> should be the module name. And maven-resolver-impl is a problem > > >> because it has two super-packages, but the module name should > > >> probably > > >> be "org.eclipse.aether.impl" with the internal package moved under > > >> impl. > > >> > > >> In summary, ignore the artifact name! Its the package name that > > >> matters when defining the module name. > > >> > > >> Stephen > > >> > > >> On 27 May 2017 at 17:43, Robert Scholte wrote: > > >>> There's no experience with this yet. > > >>> > > >>> Stephen Colebourne has written to related blogs: module naming[1] > > >>> and > > >>> modules are not artifacts[2] > > >>> which might suggest that "api" should not be added. > > >>> I do understand the addition of "api". And to make it worse, this is > > >>> probably the most important artifact needing a module name
Re: Releasing Maven Resolver 1.1.0
commit proposed in a new branch and generated site from this branch published for review: - aggregated javadoc https://maven.apache.org/resolver-archives/resolver-LATEST/apidocs/index.html - API javadoc https://maven.apache.org/resolver-archives/resolver-LATEST/maven-resolver-api/ apidocs/index.html - Implementation javadoc http://maven.apache.org/resolver-archives/resolver-LATEST/maven-resolver-impl/ apidocs/index.html Regards, Hervé Le mardi 30 mai 2017, 01:05:24 CEST Hervé BOUTEMY a écrit : > another complementary reaction while reviewing consistency between java > package names and modules names: perhaps we should change > org.apache.maven.resolver.testutil > to org.apache.maven.resolver.internal.test.util > > Regards, > > Hervé > > Le mardi 30 mai 2017, 00:50:51 CEST Hervé BOUTEMY a écrit : > > one associated question I had in mind: how do we document to end users > > what > > are the module names? Should we add a report to MPIR? And how could this > > report work, particularly on Automatic Module Name? > > > > I'm torn on choosing module name for this component: I think I understand > > Stephen's logic. > > Whatever name we choose, there will be a hard step when we move out of > > org.eclipse.aether package in Maven Resolver 2.0. > > > > I'm not sure choosing org.eclipse.aether.* module names for Maven Resolver > > 1.x and org.apache.maven.resolver.* module names for Maven Resolver 2.x > > will ease the transition: whatever the choice, the tricky history of this > > component will make its references tricky. > > > > While we do the trick at Maven artifact coords level, being consistent on > > the same trick at module names. > > One idea: perhaps adding the module names in consolidated javadoc groups > > [1] may be a solution to document the hard choice. > > > > Regards, > > > > Hervé > > > > [1] http://maven.apache.org/resolver-archives/resolver-LATEST/apidocs/ > > index.html > > > > Le lundi 29 mai 2017, 21:42:11 CEST Stephen Colebourne a écrit : > > > Well, you have my opinion. I don't think there is an exemption here > > > just because the component has a tricky history, and I personally > > > think that any exemption for the package name necessarily applies to > > > the module name (since it is now generally agreed that the module name > > > derives from the package name). > > > > > > Doing otherwise will end up being confusing as more and more people > > > name modules after super-packages. > > > > > > Stephen > > > > > > On 29 May 2017 at 18:46, Robert Scholte wrote: > > > > This makes it an interesting case :) > > > > > > > > In short: the name "Aether" is owned by Eclipse and we are not allowed > > > > to > > > > use it. > > > > However, we are allowed to use these packages for compatibility > > > > reasons > > > > as > > > > long as needed. > > > > > > > > Module names are not part of this compatibility requirement, so we > > > > shouldn't use the name Aether here. > > > > Will there be an org.apache.maven.resolver package-based > > > > implementation? > > > > Not sure, but could very well be. > > > > > > > > Based on that I think we're now using the correct/preferred module > > > > names. > > > > > > > > thanks, > > > > Robert > > > > > > > > > > > > On Mon, 29 May 2017 18:55:53 +0200, Stephen Colebourne > > > > > > > > wrote: > > > >> The module name should in almost all cases be the super-package of > > > >> the > > > >> project. > > > >> Don't use underscores in the module name unless they are also used in > > > >> the package name. > > > >> > > > >> If the super-package is "org.apache.maven.resolver.api" then that is > > > >> what the module name should be. > > > >> But if the super-package is "org.apache.maven.resolver" then that is > > > >> what the module name should be. > > > >> > > > >> ie. don't add ".api" just because that is what the artifact is > > > >> called. > > > >> Artifact name != module name. > > > >> > > > >> However, when I look at the source tree, it seems that the > > > >> super-package name is "org.eclipse.aether". If I'm right, then that > > > >> should be the module name. And maven-resolver-impl is a problem > > > >> because it has two super-packages, but the module name should > > > >> probably > > > >> be "org.eclipse.aether.impl" with the internal package moved under > > > >> impl. > > > >> > > > >> In summary, ignore the artifact name! Its the package name that > > > >> matters when defining the module name. > > > >> > > > >> Stephen > > > >> > > > >> On 27 May 2017 at 17:43, Robert Scholte wrote: > > > >>> There's no experience with this yet. > > > >>> > > > >>> Stephen Colebourne has written to related blogs: module naming[1] > > > >>> and > > > >>> modules are not artifacts[2] > > > >>> which might suggest that "api" should not be added. > > > >>> I do understand the addition of "api". And to make it worse, this is > > > >>> probably the most important artifact needing a module name. In most > > > >>> cases > > > >>> devel
Re: Releasing Maven Resolver 1.1.0
another complementary reaction while reviewing consistency between java package names and modules names: perhaps we should change org.apache.maven.resolver.testutil to org.apache.maven.resolver.internal.test.util Regards, Hervé Le mardi 30 mai 2017, 00:50:51 CEST Hervé BOUTEMY a écrit : > one associated question I had in mind: how do we document to end users what > are the module names? Should we add a report to MPIR? And how could this > report work, particularly on Automatic Module Name? > > I'm torn on choosing module name for this component: I think I understand > Stephen's logic. > Whatever name we choose, there will be a hard step when we move out of > org.eclipse.aether package in Maven Resolver 2.0. > > I'm not sure choosing org.eclipse.aether.* module names for Maven Resolver > 1.x and org.apache.maven.resolver.* module names for Maven Resolver 2.x > will ease the transition: whatever the choice, the tricky history of this > component will make its references tricky. > > While we do the trick at Maven artifact coords level, being consistent on > the same trick at module names. > One idea: perhaps adding the module names in consolidated javadoc groups [1] > may be a solution to document the hard choice. > > Regards, > > Hervé > > [1] http://maven.apache.org/resolver-archives/resolver-LATEST/apidocs/ > index.html > > Le lundi 29 mai 2017, 21:42:11 CEST Stephen Colebourne a écrit : > > Well, you have my opinion. I don't think there is an exemption here > > just because the component has a tricky history, and I personally > > think that any exemption for the package name necessarily applies to > > the module name (since it is now generally agreed that the module name > > derives from the package name). > > > > Doing otherwise will end up being confusing as more and more people > > name modules after super-packages. > > > > Stephen > > > > On 29 May 2017 at 18:46, Robert Scholte wrote: > > > This makes it an interesting case :) > > > > > > In short: the name "Aether" is owned by Eclipse and we are not allowed > > > to > > > use it. > > > However, we are allowed to use these packages for compatibility reasons > > > as > > > long as needed. > > > > > > Module names are not part of this compatibility requirement, so we > > > shouldn't use the name Aether here. > > > Will there be an org.apache.maven.resolver package-based implementation? > > > Not sure, but could very well be. > > > > > > Based on that I think we're now using the correct/preferred module > > > names. > > > > > > thanks, > > > Robert > > > > > > > > > On Mon, 29 May 2017 18:55:53 +0200, Stephen Colebourne > > > > > > wrote: > > >> The module name should in almost all cases be the super-package of the > > >> project. > > >> Don't use underscores in the module name unless they are also used in > > >> the package name. > > >> > > >> If the super-package is "org.apache.maven.resolver.api" then that is > > >> what the module name should be. > > >> But if the super-package is "org.apache.maven.resolver" then that is > > >> what the module name should be. > > >> > > >> ie. don't add ".api" just because that is what the artifact is called. > > >> Artifact name != module name. > > >> > > >> However, when I look at the source tree, it seems that the > > >> super-package name is "org.eclipse.aether". If I'm right, then that > > >> should be the module name. And maven-resolver-impl is a problem > > >> because it has two super-packages, but the module name should probably > > >> be "org.eclipse.aether.impl" with the internal package moved under > > >> impl. > > >> > > >> In summary, ignore the artifact name! Its the package name that > > >> matters when defining the module name. > > >> > > >> Stephen > > >> > > >> On 27 May 2017 at 17:43, Robert Scholte wrote: > > >>> There's no experience with this yet. > > >>> > > >>> Stephen Colebourne has written to related blogs: module naming[1] and > > >>> modules are not artifacts[2] > > >>> which might suggest that "api" should not be added. > > >>> I do understand the addition of "api". And to make it worse, this is > > >>> probably the most important artifact needing a module name. In most > > >>> cases > > >>> developers will only need this one: frameworks will handle the > > >>> implementation part. :) > > >>> > > >>> Robert > > >>> > > >>> [1] http://blog.joda.org/2017/04/java-se-9-jpms-module-naming.html > > >>> [2] > > >>> > > >>> http://blog.joda.org/2017/04/java-se-9-jpms-modules-are-not-artifacts. > > >>> ht > > >>> ml > > >>> > > >>> > > >>> On Sat, 27 May 2017 17:48:24 +0200, Hervé BOUTEMY > > >>> > > >>> > > >>> wrote: > > second option committed in another branch: > > > > option 1: > > http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/d1724eb7 > > > > option 2: > > http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/17f804d7 > > > > The only part that I'm not sure in option 2 is > > org.
Re: Releasing Maven Resolver 1.1.0
one associated question I had in mind: how do we document to end users what are the module names? Should we add a report to MPIR? And how could this report work, particularly on Automatic Module Name? I'm torn on choosing module name for this component: I think I understand Stephen's logic. Whatever name we choose, there will be a hard step when we move out of org.eclipse.aether package in Maven Resolver 2.0. I'm not sure choosing org.eclipse.aether.* module names for Maven Resolver 1.x and org.apache.maven.resolver.* module names for Maven Resolver 2.x will ease the transition: whatever the choice, the tricky history of this component will make its references tricky. While we do the trick at Maven artifact coords level, being consistent on the same trick at module names. One idea: perhaps adding the module names in consolidated javadoc groups [1] may be a solution to document the hard choice. Regards, Hervé [1] http://maven.apache.org/resolver-archives/resolver-LATEST/apidocs/ index.html Le lundi 29 mai 2017, 21:42:11 CEST Stephen Colebourne a écrit : > Well, you have my opinion. I don't think there is an exemption here > just because the component has a tricky history, and I personally > think that any exemption for the package name necessarily applies to > the module name (since it is now generally agreed that the module name > derives from the package name). > > Doing otherwise will end up being confusing as more and more people > name modules after super-packages. > > Stephen > > On 29 May 2017 at 18:46, Robert Scholte wrote: > > This makes it an interesting case :) > > > > In short: the name "Aether" is owned by Eclipse and we are not allowed to > > use it. > > However, we are allowed to use these packages for compatibility reasons as > > long as needed. > > > > Module names are not part of this compatibility requirement, so we > > shouldn't use the name Aether here. > > Will there be an org.apache.maven.resolver package-based implementation? > > Not sure, but could very well be. > > > > Based on that I think we're now using the correct/preferred module names. > > > > thanks, > > Robert > > > > > > On Mon, 29 May 2017 18:55:53 +0200, Stephen Colebourne > > > > wrote: > >> The module name should in almost all cases be the super-package of the > >> project. > >> Don't use underscores in the module name unless they are also used in > >> the package name. > >> > >> If the super-package is "org.apache.maven.resolver.api" then that is > >> what the module name should be. > >> But if the super-package is "org.apache.maven.resolver" then that is > >> what the module name should be. > >> > >> ie. don't add ".api" just because that is what the artifact is called. > >> Artifact name != module name. > >> > >> However, when I look at the source tree, it seems that the > >> super-package name is "org.eclipse.aether". If I'm right, then that > >> should be the module name. And maven-resolver-impl is a problem > >> because it has two super-packages, but the module name should probably > >> be "org.eclipse.aether.impl" with the internal package moved under > >> impl. > >> > >> In summary, ignore the artifact name! Its the package name that > >> matters when defining the module name. > >> > >> Stephen > >> > >> On 27 May 2017 at 17:43, Robert Scholte wrote: > >>> There's no experience with this yet. > >>> > >>> Stephen Colebourne has written to related blogs: module naming[1] and > >>> modules are not artifacts[2] > >>> which might suggest that "api" should not be added. > >>> I do understand the addition of "api". And to make it worse, this is > >>> probably the most important artifact needing a module name. In most > >>> cases > >>> developers will only need this one: frameworks will handle the > >>> implementation part. :) > >>> > >>> Robert > >>> > >>> [1] http://blog.joda.org/2017/04/java-se-9-jpms-module-naming.html > >>> [2] > >>> > >>> http://blog.joda.org/2017/04/java-se-9-jpms-modules-are-not-artifacts.ht > >>> ml > >>> > >>> > >>> On Sat, 27 May 2017 17:48:24 +0200, Hervé BOUTEMY > >>> > >>> > >>> wrote: > second option committed in another branch: > > option 1: > http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/d1724eb7 > > option 2: > http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/17f804d7 > > The only part that I'm not sure in option 2 is > org.apache.maven.resolver.api > > org.apache.maven.resolver: is it better to be explicit on the api or > implicit? > > Regards, > > Hervé > > Le samedi 27 mai 2017 17:37:03 CEST, vous avez écrit : > > I think I would change the following 2: > > > > org.apache.maven.resolver.connector_basic > > > org.apache.maven.resolver.connector.basic (in line with transport) > > org.apache.maven.resolver.test_util > > > org.apache.maven.resolver.testutil > > > > it's a matter of taste:
Re: Releasing Maven Resolver 1.1.0
Well, you have my opinion. I don't think there is an exemption here just because the component has a tricky history, and I personally think that any exemption for the package name necessarily applies to the module name (since it is now generally agreed that the module name derives from the package name). Doing otherwise will end up being confusing as more and more people name modules after super-packages. Stephen On 29 May 2017 at 18:46, Robert Scholte wrote: > This makes it an interesting case :) > > In short: the name "Aether" is owned by Eclipse and we are not allowed to > use it. > However, we are allowed to use these packages for compatibility reasons as > long as needed. > > Module names are not part of this compatibility requirement, so we shouldn't > use the name Aether here. > Will there be an org.apache.maven.resolver package-based implementation? Not > sure, but could very well be. > > Based on that I think we're now using the correct/preferred module names. > > thanks, > Robert > > > On Mon, 29 May 2017 18:55:53 +0200, Stephen Colebourne > wrote: > >> The module name should in almost all cases be the super-package of the >> project. >> Don't use underscores in the module name unless they are also used in >> the package name. >> >> If the super-package is "org.apache.maven.resolver.api" then that is >> what the module name should be. >> But if the super-package is "org.apache.maven.resolver" then that is >> what the module name should be. >> >> ie. don't add ".api" just because that is what the artifact is called. >> Artifact name != module name. >> >> However, when I look at the source tree, it seems that the >> super-package name is "org.eclipse.aether". If I'm right, then that >> should be the module name. And maven-resolver-impl is a problem >> because it has two super-packages, but the module name should probably >> be "org.eclipse.aether.impl" with the internal package moved under >> impl. >> >> In summary, ignore the artifact name! Its the package name that >> matters when defining the module name. >> >> Stephen >> >> >> On 27 May 2017 at 17:43, Robert Scholte wrote: >>> >>> There's no experience with this yet. >>> >>> Stephen Colebourne has written to related blogs: module naming[1] and >>> modules are not artifacts[2] >>> which might suggest that "api" should not be added. >>> I do understand the addition of "api". And to make it worse, this is >>> probably the most important artifact needing a module name. In most cases >>> developers will only need this one: frameworks will handle the >>> implementation part. :) >>> >>> Robert >>> >>> [1] http://blog.joda.org/2017/04/java-se-9-jpms-module-naming.html >>> [2] >>> >>> http://blog.joda.org/2017/04/java-se-9-jpms-modules-are-not-artifacts.html >>> >>> >>> On Sat, 27 May 2017 17:48:24 +0200, Hervé BOUTEMY >>> wrote: >>> second option committed in another branch: option 1: http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/d1724eb7 option 2: http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/17f804d7 The only part that I'm not sure in option 2 is org.apache.maven.resolver.api > org.apache.maven.resolver: is it better to be explicit on the api or implicit? Regards, Hervé Le samedi 27 mai 2017 17:37:03 CEST, vous avez écrit : > > > I think I would change the following 2: > > org.apache.maven.resolver.connector_basic > > org.apache.maven.resolver.connector.basic (in line with transport) > org.apache.maven.resolver.test_util > > org.apache.maven.resolver.testutil > > it's a matter of taste: the underscores look kind of weird, but that's > probably caused because we've never used them as package names. > > And wondering if "api" should be changed, since this is the > access-module > and we don't use api in our packages: > org.apache.maven.resolver.api > org.apache.maven.resolver > > Using a property makes it easier to configure the maven-jar-plugin, but > it > also makes these constants turn into variables, i.e. you could set them > via system properties or cmdline args. > If only we supported something like > > > ${project.properties["AutomaticModuleName"]} ic-Module-Name> > > for the rest it's looking good. > > thanks > Robert > > > On Sat, 27 May 2017 17:20:15 +0200, Hervé BOUTEMY > > > wrote: > > please review and second if you think it's ok: > > http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/d1724eb7 > > > > Regards, > > > > Hervé > > > > Le samedi 27 mai 2017, 13:24:47 CEST Hervé BOUTEMY a écrit : > >> he he, Java 9 is really coming, with associated real world > >> questions. > >> > >> Maven Artifact Resolver is one of rare Maven components that has a > >> chance to > >> become a collection Java 9 modules, si
Re: Releasing Maven Resolver 1.1.0
This makes it an interesting case :) In short: the name "Aether" is owned by Eclipse and we are not allowed to use it. However, we are allowed to use these packages for compatibility reasons as long as needed. Module names are not part of this compatibility requirement, so we shouldn't use the name Aether here. Will there be an org.apache.maven.resolver package-based implementation? Not sure, but could very well be. Based on that I think we're now using the correct/preferred module names. thanks, Robert On Mon, 29 May 2017 18:55:53 +0200, Stephen Colebourne wrote: The module name should in almost all cases be the super-package of the project. Don't use underscores in the module name unless they are also used in the package name. If the super-package is "org.apache.maven.resolver.api" then that is what the module name should be. But if the super-package is "org.apache.maven.resolver" then that is what the module name should be. ie. don't add ".api" just because that is what the artifact is called. Artifact name != module name. However, when I look at the source tree, it seems that the super-package name is "org.eclipse.aether". If I'm right, then that should be the module name. And maven-resolver-impl is a problem because it has two super-packages, but the module name should probably be "org.eclipse.aether.impl" with the internal package moved under impl. In summary, ignore the artifact name! Its the package name that matters when defining the module name. Stephen On 27 May 2017 at 17:43, Robert Scholte wrote: There's no experience with this yet. Stephen Colebourne has written to related blogs: module naming[1] and modules are not artifacts[2] which might suggest that "api" should not be added. I do understand the addition of "api". And to make it worse, this is probably the most important artifact needing a module name. In most cases developers will only need this one: frameworks will handle the implementation part. :) Robert [1] http://blog.joda.org/2017/04/java-se-9-jpms-module-naming.html [2] http://blog.joda.org/2017/04/java-se-9-jpms-modules-are-not-artifacts.html On Sat, 27 May 2017 17:48:24 +0200, Hervé BOUTEMY wrote: second option committed in another branch: option 1: http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/d1724eb7 option 2: http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/17f804d7 The only part that I'm not sure in option 2 is org.apache.maven.resolver.api > org.apache.maven.resolver: is it better to be explicit on the api or implicit? Regards, Hervé Le samedi 27 mai 2017 17:37:03 CEST, vous avez écrit : I think I would change the following 2: org.apache.maven.resolver.connector_basic > org.apache.maven.resolver.connector.basic (in line with transport) org.apache.maven.resolver.test_util > org.apache.maven.resolver.testutil it's a matter of taste: the underscores look kind of weird, but that's probably caused because we've never used them as package names. And wondering if "api" should be changed, since this is the access-module and we don't use api in our packages: org.apache.maven.resolver.api > org.apache.maven.resolver Using a property makes it easier to configure the maven-jar-plugin, but it also makes these constants turn into variables, i.e. you could set them via system properties or cmdline args. If only we supported something like ${project.properties["AutomaticModuleName"]} for the rest it's looking good. thanks Robert On Sat, 27 May 2017 17:20:15 +0200, Hervé BOUTEMY wrote: > please review and second if you think it's ok: > http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/d1724eb7 > > Regards, > > Hervé > > Le samedi 27 mai 2017, 13:24:47 CEST Hervé BOUTEMY a écrit : >> he he, Java 9 is really coming, with associated real world questions. >> >> Maven Artifact Resolver is one of rare Maven components that has a >> chance to >> become a collection Java 9 modules, since it was written quite >> recently >> and >> is pure new code as a result of Maven 3 refactoring, then does not >> have >> shared package names issues we have with Maven core itself. >> >> And since it is expected to be a lib for easy embedding of artifact >> resolution, making it a collection of Java 9 modules would be not only >> a >> great opportunity to test Java 9 modules, but it could be useful for >> people >> using it. >> >> Then I'm highly in favor of trying. >> >> Adding Automatic-Module-Name to the MANIFEST.MF looks feasible right >> now, >> without waiting much: I'm pretty sure module names will be obvious, >> and >> much >> better if we define them instead of waiting for automatic names. Let's >> start! MRESOLVER-26 created [1] >> >> Then there is the question of making real modules: is it feasible >> right >> now? >> Or would we need a delay to tweak the module descriptors? And that >> would >> mean that we need Java 9 to build, even if the generated bytecode >> remai
Re: Releasing Maven Resolver 1.1.0
The module name should in almost all cases be the super-package of the project. Don't use underscores in the module name unless they are also used in the package name. If the super-package is "org.apache.maven.resolver.api" then that is what the module name should be. But if the super-package is "org.apache.maven.resolver" then that is what the module name should be. ie. don't add ".api" just because that is what the artifact is called. Artifact name != module name. However, when I look at the source tree, it seems that the super-package name is "org.eclipse.aether". If I'm right, then that should be the module name. And maven-resolver-impl is a problem because it has two super-packages, but the module name should probably be "org.eclipse.aether.impl" with the internal package moved under impl. In summary, ignore the artifact name! Its the package name that matters when defining the module name. Stephen On 27 May 2017 at 17:43, Robert Scholte wrote: > There's no experience with this yet. > > Stephen Colebourne has written to related blogs: module naming[1] and > modules are not artifacts[2] > which might suggest that "api" should not be added. > I do understand the addition of "api". And to make it worse, this is > probably the most important artifact needing a module name. In most cases > developers will only need this one: frameworks will handle the > implementation part. :) > > Robert > > [1] http://blog.joda.org/2017/04/java-se-9-jpms-module-naming.html > [2] > http://blog.joda.org/2017/04/java-se-9-jpms-modules-are-not-artifacts.html > > > On Sat, 27 May 2017 17:48:24 +0200, Hervé BOUTEMY > wrote: > >> second option committed in another branch: >> >> option 1: >> http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/d1724eb7 >> >> option 2: >> http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/17f804d7 >> >> The only part that I'm not sure in option 2 is >> org.apache.maven.resolver.api > >> org.apache.maven.resolver: is it better to be explicit on the api or >> implicit? >> >> Regards, >> >> Hervé >> >> Le samedi 27 mai 2017 17:37:03 CEST, vous avez écrit : >>> >>> I think I would change the following 2: >>> >>> org.apache.maven.resolver.connector_basic > >>> org.apache.maven.resolver.connector.basic (in line with transport) >>> org.apache.maven.resolver.test_util > org.apache.maven.resolver.testutil >>> >>> it's a matter of taste: the underscores look kind of weird, but that's >>> probably caused because we've never used them as package names. >>> >>> And wondering if "api" should be changed, since this is the access-module >>> and we don't use api in our packages: >>> org.apache.maven.resolver.api > org.apache.maven.resolver >>> >>> Using a property makes it easier to configure the maven-jar-plugin, but >>> it >>> also makes these constants turn into variables, i.e. you could set them >>> via system properties or cmdline args. >>> If only we supported something like >>> >>> ${project.properties["AutomaticModuleName"]}>> ic-Module-Name> >>> >>> for the rest it's looking good. >>> >>> thanks >>> Robert >>> >>> >>> On Sat, 27 May 2017 17:20:15 +0200, Hervé BOUTEMY >>> >>> wrote: >>> > please review and second if you think it's ok: >>> > http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/d1724eb7 >>> > >>> > Regards, >>> > >>> > Hervé >>> > >>> > Le samedi 27 mai 2017, 13:24:47 CEST Hervé BOUTEMY a écrit : >>> >> he he, Java 9 is really coming, with associated real world questions. >>> >> >>> >> Maven Artifact Resolver is one of rare Maven components that has a >>> >> chance to >>> >> become a collection Java 9 modules, since it was written quite >>> >> recently >>> >> and >>> >> is pure new code as a result of Maven 3 refactoring, then does not >>> >> have >>> >> shared package names issues we have with Maven core itself. >>> >> >>> >> And since it is expected to be a lib for easy embedding of artifact >>> >> resolution, making it a collection of Java 9 modules would be not only >>> >> a >>> >> great opportunity to test Java 9 modules, but it could be useful for >>> >> people >>> >> using it. >>> >> >>> >> Then I'm highly in favor of trying. >>> >> >>> >> Adding Automatic-Module-Name to the MANIFEST.MF looks feasible right >>> >> now, >>> >> without waiting much: I'm pretty sure module names will be obvious, >>> >> and >>> >> much >>> >> better if we define them instead of waiting for automatic names. Let's >>> >> start! MRESOLVER-26 created [1] >>> >> >>> >> Then there is the question of making real modules: is it feasible >>> >> right >>> >> now? >>> >> Or would we need a delay to tweak the module descriptors? And that >>> >> would >>> >> mean that we need Java 9 to build, even if the generated bytecode >>> >> remains >>> >> Java 7 compatible, isn't it? Is Maven tooling ready to it? >>> >> MRESOLVER-27 created to track the issue [2], but I'm not sure this is >>> >> the >>> >> right time to do this job, but for the next release after this 1.1.0 >>> >> >>> >> Regar
Re: Releasing Maven Resolver 1.1.0
done Le dimanche 28 mai 2017, 19:54:19 CEST Michael Osipov a écrit : > Am 2017-05-28 um 17:38 schrieb Hervé BOUTEMY: > > Michael, > > > > is it ok for you now? > > I prefer option 2. Go ahead with it. > > > Le dimanche 28 mai 2017, 11:16:58 CEST Arnaud Héritier a écrit : > >> Let's go for option 2 > >> > >> Le dim. 28 mai 2017 à 12:44, Robert Scholte a écrit : > >>> On behalf of the expert group I can confirm we agreed on this solution. > >>> I don't see any reason why this would change as this topic is marked as > >>> resolved. > >>> And I think it is a good sign, for some reason there is/was this rumor > >>> that Maven doesn't run on J9. > >>> > >>> I second option 2. > >>> > >>> thanks, > >>> Robert > >>> > >>> On Sun, 28 May 2017 11:15:00 +0200, Michael Osipov > >>> > >>> wrote: > Am 2017-05-28 um 09:43 schrieb Hervé BOUTEMY: > > are there seconders for > > http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/17f804d7 > > (aka "option 2")? > > I'd completely leave it off to 1.x until the expect group with Mark > Reinhold has agreed on the disputed points. > > I don't see a reason to put any effort into a system which is still in > constant flux. > > > Le samedi 27 mai 2017, 19:05:27 CEST Hervé BOUTEMY a écrit : > >> good links > >> yes, with this in mind, "api" is required for artifactId but should > >> not be > >> added to module name: good catch, and good experience to share > >> because > >> that > >> was not so obvious > >> > >> Regards, > >> > >> Hervé > >> > >> Le samedi 27 mai 2017, 18:43:22 CEST Robert Scholte a écrit : > >>> There's no experience with this yet. > >>> > >>> Stephen Colebourne has written to related blogs: module naming[1] > >>> and > >>> modules are not artifacts[2] > >>> which might suggest that "api" should not be added. > >>> I do understand the addition of "api". And to make it worse, this is > >>> probably the most important artifact needing a module name. In most > >>> cases > >>> developers will only need this one: frameworks will handle the > >>> implementation part. :) > >>> > >>> Robert > >>> > >>> [1] http://blog.joda.org/2017/04/java-se-9-jpms-module-naming.html > >>> [2] > >>> > >>> http://blog.joda.org/2017/04/java-se-9-jpms-modules-are-not-artifacts.ht > >>> ml > >>> > >>> On Sat, 27 May 2017 17:48:24 +0200, Hervé BOUTEMY > >>> > >>> > >>> wrote: > second option committed in another branch: > >>> > option 1: > >>> http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/d1724eb7 > >>> > option 2: > >>> http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/17f804d7 > >>> > The only part that I'm not sure in option 2 is > org.apache.maven.resolver.api > > org.apache.maven.resolver: is it better to be explicit on the api > or > implicit? > > Regards, > > Hervé > > Le samedi 27 mai 2017 17:37:03 CEST, vous avez écrit : > > I think I would change the following 2: > > > > org.apache.maven.resolver.connector_basic > > > org.apache.maven.resolver.connector.basic (in line with transport) > > org.apache.maven.resolver.test_util > > > org.apache.maven.resolver.testutil > > > > it's a matter of taste: the underscores look kind of weird, but > > that's > > probably caused because we've never used them as package names. > > > > And wondering if "api" should be changed, since this is the > > access-module > > and we don't use api in our packages: > > org.apache.maven.resolver.api > org.apache.maven.resolver > > > > Using a property makes it easier to configure the > > maven-jar-plugin, > > but > > it > > also makes these constants turn into variables, i.e. you could set > > them > > via system properties or cmdline args. > > If only we supported something like > >>> > >>> ${project.properties["AutomaticModuleName"]} >>> > > to > > mat ic-Module-Name> > > > > for the rest it's looking good. > > > > thanks > > Robert > > > > > > On Sat, 27 May 2017 17:20:15 +0200, Hervé BOUTEMY > > > > > > wrote: > >> please review and second if you think it's ok: > >>> http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/d1724eb7 > >>> > >> Regards, > >> > >> Hervé > >> > >> Le samedi 27 mai 2017, 13:24:47 CEST Hervé BOUTEMY a écrit : > >>> he he, Java 9 is really coming, with associated real world > >>> questions. > >>> > >>> Maven Artifact Reso
Re: Releasing Maven Resolver 1.1.0
Am 2017-05-28 um 17:38 schrieb Hervé BOUTEMY: Michael, is it ok for you now? I prefer option 2. Go ahead with it. Le dimanche 28 mai 2017, 11:16:58 CEST Arnaud Héritier a écrit : Let's go for option 2 Le dim. 28 mai 2017 à 12:44, Robert Scholte a écrit : On behalf of the expert group I can confirm we agreed on this solution. I don't see any reason why this would change as this topic is marked as resolved. And I think it is a good sign, for some reason there is/was this rumor that Maven doesn't run on J9. I second option 2. thanks, Robert On Sun, 28 May 2017 11:15:00 +0200, Michael Osipov wrote: Am 2017-05-28 um 09:43 schrieb Hervé BOUTEMY: are there seconders for http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/17f804d7 (aka "option 2")? I'd completely leave it off to 1.x until the expect group with Mark Reinhold has agreed on the disputed points. I don't see a reason to put any effort into a system which is still in constant flux. Le samedi 27 mai 2017, 19:05:27 CEST Hervé BOUTEMY a écrit : good links yes, with this in mind, "api" is required for artifactId but should not be added to module name: good catch, and good experience to share because that was not so obvious Regards, Hervé Le samedi 27 mai 2017, 18:43:22 CEST Robert Scholte a écrit : There's no experience with this yet. Stephen Colebourne has written to related blogs: module naming[1] and modules are not artifacts[2] which might suggest that "api" should not be added. I do understand the addition of "api". And to make it worse, this is probably the most important artifact needing a module name. In most cases developers will only need this one: frameworks will handle the implementation part. :) Robert [1] http://blog.joda.org/2017/04/java-se-9-jpms-module-naming.html [2] http://blog.joda.org/2017/04/java-se-9-jpms-modules-are-not-artifacts.html On Sat, 27 May 2017 17:48:24 +0200, Hervé BOUTEMY wrote: second option committed in another branch: option 1: http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/d1724eb7 option 2: http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/17f804d7 The only part that I'm not sure in option 2 is org.apache.maven.resolver.api > org.apache.maven.resolver: is it better to be explicit on the api or implicit? Regards, Hervé Le samedi 27 mai 2017 17:37:03 CEST, vous avez écrit : I think I would change the following 2: org.apache.maven.resolver.connector_basic > org.apache.maven.resolver.connector.basic (in line with transport) org.apache.maven.resolver.test_util > org.apache.maven.resolver.testutil it's a matter of taste: the underscores look kind of weird, but that's probably caused because we've never used them as package names. And wondering if "api" should be changed, since this is the access-module and we don't use api in our packages: org.apache.maven.resolver.api > org.apache.maven.resolver Using a property makes it easier to configure the maven-jar-plugin, but it also makes these constants turn into variables, i.e. you could set them via system properties or cmdline args. If only we supported something like ${project.properties["AutomaticModuleName"]} to mat ic-Module-Name> for the rest it's looking good. thanks Robert On Sat, 27 May 2017 17:20:15 +0200, Hervé BOUTEMY wrote: please review and second if you think it's ok: http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/d1724eb7 Regards, Hervé Le samedi 27 mai 2017, 13:24:47 CEST Hervé BOUTEMY a écrit : he he, Java 9 is really coming, with associated real world questions. Maven Artifact Resolver is one of rare Maven components that has a chance to become a collection Java 9 modules, since it was written quite recently and is pure new code as a result of Maven 3 refactoring, then does not have shared package names issues we have with Maven core itself. And since it is expected to be a lib for easy embedding of artifact resolution, making it a collection of Java 9 modules would be not only a great opportunity to test Java 9 modules, but it could be useful for people using it. Then I'm highly in favor of trying. Adding Automatic-Module-Name to the MANIFEST.MF looks feasible right now, without waiting much: I'm pretty sure module names will be obvious, and much better if we define them instead of waiting for automatic names. Let's start! MRESOLVER-26 created [1] Then there is the question of making real modules: is it feasible right now? Or would we need a delay to tweak the module descriptors? And that would mean that we need Java 9 to build, even if the generated bytecode remains Java 7 compatible, isn't it? Is Maven tooling ready to it? MRESOLVER-27 created to track the issue [2], but I'm not sure this is the right time to do this job, but for the next release after this 1.1.0 Regards, Hervé [1] https://issues.apache.org/jira/browse/MRESOLVER-26 [2] https://issues.apache.org/jira/browse/MRESOLVER-27 Le samedi 27 mai
Re: Releasing Maven Resolver 1.1.0
Michael, is it ok for you now? Regards, Hervé Le dimanche 28 mai 2017, 11:16:58 CEST Arnaud Héritier a écrit : > Let's go for option 2 > > Le dim. 28 mai 2017 à 12:44, Robert Scholte a écrit : > > On behalf of the expert group I can confirm we agreed on this solution. > > I don't see any reason why this would change as this topic is marked as > > resolved. > > And I think it is a good sign, for some reason there is/was this rumor > > that Maven doesn't run on J9. > > > > I second option 2. > > > > thanks, > > Robert > > > > On Sun, 28 May 2017 11:15:00 +0200, Michael Osipov > > > > wrote: > > > Am 2017-05-28 um 09:43 schrieb Hervé BOUTEMY: > > >> are there seconders for > > >> http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/17f804d7 > > >> (aka "option 2")? > > > > > > I'd completely leave it off to 1.x until the expect group with Mark > > > Reinhold has agreed on the disputed points. > > > > > > I don't see a reason to put any effort into a system which is still in > > > constant flux. > > > > > >> Le samedi 27 mai 2017, 19:05:27 CEST Hervé BOUTEMY a écrit : > > >>> good links > > >>> yes, with this in mind, "api" is required for artifactId but should > > >>> not be > > >>> added to module name: good catch, and good experience to share because > > >>> that > > >>> was not so obvious > > >>> > > >>> Regards, > > >>> > > >>> Hervé > > >>> > > >>> Le samedi 27 mai 2017, 18:43:22 CEST Robert Scholte a écrit : > > There's no experience with this yet. > > > > Stephen Colebourne has written to related blogs: module naming[1] and > > modules are not artifacts[2] > > which might suggest that "api" should not be added. > > I do understand the addition of "api". And to make it worse, this is > > probably the most important artifact needing a module name. In most > > cases > > developers will only need this one: frameworks will handle the > > implementation part. :) > > > > Robert > > > > [1] http://blog.joda.org/2017/04/java-se-9-jpms-module-naming.html > > [2] > > > > http://blog.joda.org/2017/04/java-se-9-jpms-modules-are-not-artifacts.html > > > > On Sat, 27 May 2017 17:48:24 +0200, Hervé BOUTEMY > > > > > > wrote: > > > second option committed in another branch: > > > > > option 1: > > http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/d1724eb7 > > > > > option 2: > > http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/17f804d7 > > > > > The only part that I'm not sure in option 2 is > > > org.apache.maven.resolver.api > > > > org.apache.maven.resolver: is it better to be explicit on the api or > > > implicit? > > > > > > Regards, > > > > > > Hervé > > > > > > Le samedi 27 mai 2017 17:37:03 CEST, vous avez écrit : > > >> I think I would change the following 2: > > >> > > >> org.apache.maven.resolver.connector_basic > > > >> org.apache.maven.resolver.connector.basic (in line with transport) > > >> org.apache.maven.resolver.test_util > > > >> org.apache.maven.resolver.testutil > > >> > > >> it's a matter of taste: the underscores look kind of weird, but > > >> that's > > >> probably caused because we've never used them as package names. > > >> > > >> And wondering if "api" should be changed, since this is the > > >> access-module > > >> and we don't use api in our packages: > > >> org.apache.maven.resolver.api > org.apache.maven.resolver > > >> > > >> Using a property makes it easier to configure the maven-jar-plugin, > > >> but > > >> it > > >> also makes these constants turn into variables, i.e. you could set > > >> them > > >> via system properties or cmdline args. > > >> If only we supported something like > > > > ${project.properties["AutomaticModuleName"]} > > > >> to > > >> mat ic-Module-Name> > > >> > > >> for the rest it's looking good. > > >> > > >> thanks > > >> Robert > > >> > > >> > > >> On Sat, 27 May 2017 17:20:15 +0200, Hervé BOUTEMY > > >> > > >> > > >> wrote: > > >>> please review and second if you think it's ok: > > http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/d1724eb7 > > > > >>> Regards, > > >>> > > >>> Hervé > > >>> > > >>> Le samedi 27 mai 2017, 13:24:47 CEST Hervé BOUTEMY a écrit : > > he he, Java 9 is really coming, with associated real world > > questions. > > > > Maven Artifact Resolver is one of rare Maven components that has > > a > > chance to > > become a collection Java 9 modules, since it was written quite > > >> > > >> recently > > >> > > and > > is pure new code as a result of Maven 3 refactoring, then does > > not > > >> > > >> have > > >> > > shared pa
Re: Releasing Maven Resolver 1.1.0
Let's go for option 2 Le dim. 28 mai 2017 à 12:44, Robert Scholte a écrit : > On behalf of the expert group I can confirm we agreed on this solution. > I don't see any reason why this would change as this topic is marked as > resolved. > And I think it is a good sign, for some reason there is/was this rumor > that Maven doesn't run on J9. > > I second option 2. > > thanks, > Robert > > On Sun, 28 May 2017 11:15:00 +0200, Michael Osipov > wrote: > > > Am 2017-05-28 um 09:43 schrieb Hervé BOUTEMY: > >> are there seconders for > >> http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/17f804d7 > >> (aka "option 2")? > > > > I'd completely leave it off to 1.x until the expect group with Mark > > Reinhold has agreed on the disputed points. > > > > I don't see a reason to put any effort into a system which is still in > > constant flux. > > > >> Le samedi 27 mai 2017, 19:05:27 CEST Hervé BOUTEMY a écrit : > >>> good links > >>> yes, with this in mind, "api" is required for artifactId but should > >>> not be > >>> added to module name: good catch, and good experience to share because > >>> that > >>> was not so obvious > >>> > >>> Regards, > >>> > >>> Hervé > >>> > >>> Le samedi 27 mai 2017, 18:43:22 CEST Robert Scholte a écrit : > There's no experience with this yet. > > Stephen Colebourne has written to related blogs: module naming[1] and > modules are not artifacts[2] > which might suggest that "api" should not be added. > I do understand the addition of "api". And to make it worse, this is > probably the most important artifact needing a module name. In most > cases > developers will only need this one: frameworks will handle the > implementation part. :) > > Robert > > [1] http://blog.joda.org/2017/04/java-se-9-jpms-module-naming.html > [2] > > http://blog.joda.org/2017/04/java-se-9-jpms-modules-are-not-artifacts.html > > > On Sat, 27 May 2017 17:48:24 +0200, Hervé BOUTEMY > > > wrote: > > second option committed in another branch: > > > > option 1: > > > http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/d1724eb7 > > > > option 2: > > > http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/17f804d7 > > > > The only part that I'm not sure in option 2 is > > org.apache.maven.resolver.api > > > org.apache.maven.resolver: is it better to be explicit on the api or > > implicit? > > > > Regards, > > > > Hervé > > > > Le samedi 27 mai 2017 17:37:03 CEST, vous avez écrit : > >> I think I would change the following 2: > >> > >> org.apache.maven.resolver.connector_basic > > >> org.apache.maven.resolver.connector.basic (in line with transport) > >> org.apache.maven.resolver.test_util > > >> org.apache.maven.resolver.testutil > >> > >> it's a matter of taste: the underscores look kind of weird, but > >> that's > >> probably caused because we've never used them as package names. > >> > >> And wondering if "api" should be changed, since this is the > >> access-module > >> and we don't use api in our packages: > >> org.apache.maven.resolver.api > org.apache.maven.resolver > >> > >> Using a property makes it easier to configure the maven-jar-plugin, > >> but > >> it > >> also makes these constants turn into variables, i.e. you could set > >> them > >> via system properties or cmdline args. > >> If only we supported something like > >> > ${project.properties["AutomaticModuleName"]} >> to > >> mat ic-Module-Name> > >> > >> for the rest it's looking good. > >> > >> thanks > >> Robert > >> > >> > >> On Sat, 27 May 2017 17:20:15 +0200, Hervé BOUTEMY > >> > >> > >> wrote: > >>> please review and second if you think it's ok: > >>> > http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/d1724eb7 > >>> > >>> Regards, > >>> > >>> Hervé > >>> > >>> Le samedi 27 mai 2017, 13:24:47 CEST Hervé BOUTEMY a écrit : > he he, Java 9 is really coming, with associated real world > questions. > > Maven Artifact Resolver is one of rare Maven components that has a > chance to > become a collection Java 9 modules, since it was written quite > >> > >> recently > >> > and > is pure new code as a result of Maven 3 refactoring, then does not > >> > >> have > >> > shared package names issues we have with Maven core itself. > > And since it is expected to be a lib for easy embedding of > artifact > resolution, making it a collection of Java 9 modules would be not > >> > >> only a > >> > great opportunity to test Java 9 modules, but it could be useful > for > people > using it.
Re: Releasing Maven Resolver 1.1.0
On behalf of the expert group I can confirm we agreed on this solution. I don't see any reason why this would change as this topic is marked as resolved. And I think it is a good sign, for some reason there is/was this rumor that Maven doesn't run on J9. I second option 2. thanks, Robert On Sun, 28 May 2017 11:15:00 +0200, Michael Osipov wrote: Am 2017-05-28 um 09:43 schrieb Hervé BOUTEMY: are there seconders for http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/17f804d7 (aka "option 2")? I'd completely leave it off to 1.x until the expect group with Mark Reinhold has agreed on the disputed points. I don't see a reason to put any effort into a system which is still in constant flux. Le samedi 27 mai 2017, 19:05:27 CEST Hervé BOUTEMY a écrit : good links yes, with this in mind, "api" is required for artifactId but should not be added to module name: good catch, and good experience to share because that was not so obvious Regards, Hervé Le samedi 27 mai 2017, 18:43:22 CEST Robert Scholte a écrit : There's no experience with this yet. Stephen Colebourne has written to related blogs: module naming[1] and modules are not artifacts[2] which might suggest that "api" should not be added. I do understand the addition of "api". And to make it worse, this is probably the most important artifact needing a module name. In most cases developers will only need this one: frameworks will handle the implementation part. :) Robert [1] http://blog.joda.org/2017/04/java-se-9-jpms-module-naming.html [2] http://blog.joda.org/2017/04/java-se-9-jpms-modules-are-not-artifacts.html On Sat, 27 May 2017 17:48:24 +0200, Hervé BOUTEMY wrote: second option committed in another branch: option 1: http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/d1724eb7 option 2: http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/17f804d7 The only part that I'm not sure in option 2 is org.apache.maven.resolver.api > org.apache.maven.resolver: is it better to be explicit on the api or implicit? Regards, Hervé Le samedi 27 mai 2017 17:37:03 CEST, vous avez écrit : I think I would change the following 2: org.apache.maven.resolver.connector_basic > org.apache.maven.resolver.connector.basic (in line with transport) org.apache.maven.resolver.test_util > org.apache.maven.resolver.testutil it's a matter of taste: the underscores look kind of weird, but that's probably caused because we've never used them as package names. And wondering if "api" should be changed, since this is the access-module and we don't use api in our packages: org.apache.maven.resolver.api > org.apache.maven.resolver Using a property makes it easier to configure the maven-jar-plugin, but it also makes these constants turn into variables, i.e. you could set them via system properties or cmdline args. If only we supported something like ${project.properties["AutomaticModuleName"]} for the rest it's looking good. thanks Robert On Sat, 27 May 2017 17:20:15 +0200, Hervé BOUTEMY wrote: please review and second if you think it's ok: http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/d1724eb7 Regards, Hervé Le samedi 27 mai 2017, 13:24:47 CEST Hervé BOUTEMY a écrit : he he, Java 9 is really coming, with associated real world questions. Maven Artifact Resolver is one of rare Maven components that has a chance to become a collection Java 9 modules, since it was written quite recently and is pure new code as a result of Maven 3 refactoring, then does not have shared package names issues we have with Maven core itself. And since it is expected to be a lib for easy embedding of artifact resolution, making it a collection of Java 9 modules would be not only a great opportunity to test Java 9 modules, but it could be useful for people using it. Then I'm highly in favor of trying. Adding Automatic-Module-Name to the MANIFEST.MF looks feasible right now, without waiting much: I'm pretty sure module names will be obvious, and much better if we define them instead of waiting for automatic names. Let's start! MRESOLVER-26 created [1] Then there is the question of making real modules: is it feasible right now? Or would we need a delay to tweak the module descriptors? And that would mean that we need Java 9 to build, even if the generated bytecode remains Java 7 compatible, isn't it? Is Maven tooling ready to it? MRESOLVER-27 created to track the issue [2], but I'm not sure this is the right time to do this job, but for the next release after this 1.1.0 Regards, Hervé [1] https://issues.apache.org/jira/browse/MRESOLVER-26 [2] https://issues.apache.org/jira/browse/MRESOLVER-27 Le samedi 27 mai 2017, 11:58:43 CEST Robert Scholte a écrit : Hi, I've got a question from Remi Forax if we could add Java9 module descriptors to this project. This will be one of the first which can provide such descriptors since it has no required dependencies o
Re: Releasing Maven Resolver 1.1.0
Am 2017-05-28 um 09:43 schrieb Hervé BOUTEMY: are there seconders for http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/17f804d7 (aka "option 2")? I'd completely leave it off to 1.x until the expect group with Mark Reinhold has agreed on the disputed points. I don't see a reason to put any effort into a system which is still in constant flux. Le samedi 27 mai 2017, 19:05:27 CEST Hervé BOUTEMY a écrit : good links yes, with this in mind, "api" is required for artifactId but should not be added to module name: good catch, and good experience to share because that was not so obvious Regards, Hervé Le samedi 27 mai 2017, 18:43:22 CEST Robert Scholte a écrit : There's no experience with this yet. Stephen Colebourne has written to related blogs: module naming[1] and modules are not artifacts[2] which might suggest that "api" should not be added. I do understand the addition of "api". And to make it worse, this is probably the most important artifact needing a module name. In most cases developers will only need this one: frameworks will handle the implementation part. :) Robert [1] http://blog.joda.org/2017/04/java-se-9-jpms-module-naming.html [2] http://blog.joda.org/2017/04/java-se-9-jpms-modules-are-not-artifacts.html On Sat, 27 May 2017 17:48:24 +0200, Hervé BOUTEMY wrote: second option committed in another branch: option 1: http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/d1724eb7 option 2: http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/17f804d7 The only part that I'm not sure in option 2 is org.apache.maven.resolver.api > org.apache.maven.resolver: is it better to be explicit on the api or implicit? Regards, Hervé Le samedi 27 mai 2017 17:37:03 CEST, vous avez écrit : I think I would change the following 2: org.apache.maven.resolver.connector_basic > org.apache.maven.resolver.connector.basic (in line with transport) org.apache.maven.resolver.test_util > org.apache.maven.resolver.testutil it's a matter of taste: the underscores look kind of weird, but that's probably caused because we've never used them as package names. And wondering if "api" should be changed, since this is the access-module and we don't use api in our packages: org.apache.maven.resolver.api > org.apache.maven.resolver Using a property makes it easier to configure the maven-jar-plugin, but it also makes these constants turn into variables, i.e. you could set them via system properties or cmdline args. If only we supported something like ${project.properties["AutomaticModuleName"]} for the rest it's looking good. thanks Robert On Sat, 27 May 2017 17:20:15 +0200, Hervé BOUTEMY wrote: please review and second if you think it's ok: http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/d1724eb7 Regards, Hervé Le samedi 27 mai 2017, 13:24:47 CEST Hervé BOUTEMY a écrit : he he, Java 9 is really coming, with associated real world questions. Maven Artifact Resolver is one of rare Maven components that has a chance to become a collection Java 9 modules, since it was written quite recently and is pure new code as a result of Maven 3 refactoring, then does not have shared package names issues we have with Maven core itself. And since it is expected to be a lib for easy embedding of artifact resolution, making it a collection of Java 9 modules would be not only a great opportunity to test Java 9 modules, but it could be useful for people using it. Then I'm highly in favor of trying. Adding Automatic-Module-Name to the MANIFEST.MF looks feasible right now, without waiting much: I'm pretty sure module names will be obvious, and much better if we define them instead of waiting for automatic names. Let's start! MRESOLVER-26 created [1] Then there is the question of making real modules: is it feasible right now? Or would we need a delay to tweak the module descriptors? And that would mean that we need Java 9 to build, even if the generated bytecode remains Java 7 compatible, isn't it? Is Maven tooling ready to it? MRESOLVER-27 created to track the issue [2], but I'm not sure this is the right time to do this job, but for the next release after this 1.1.0 Regards, Hervé [1] https://issues.apache.org/jira/browse/MRESOLVER-26 [2] https://issues.apache.org/jira/browse/MRESOLVER-27 Le samedi 27 mai 2017, 11:58:43 CEST Robert Scholte a écrit : Hi, I've got a question from Remi Forax if we could add Java9 module descriptors to this project. This will be one of the first which can provide such descriptors since it has no required dependencies other then its own and its package structure seems valid with the new Java9 rules. We haven't discussed this in general yet, but we have several projects which are at the bottom of the dependency tree which should provide either a module name or module descriptor when possible. Do we want to help the community by having already several libraries with a module descriptor? Or
Re: Releasing Maven Resolver 1.1.0
Am 2017-05-27 um 11:42 schrieb Hervé BOUTEMY: Hi, No objection from me, thanks for keeping the ball rolling. I tried to improve documentation by adding some useful links to other related components [1]: I think the current state is better and ok for a release. One key question now is about Aether wiki content [2]: should we copy it? In a wiki or in components sources? I suppose wiki source format is supported by Doxia, then it could be imported quite easily in sources. The wiki docs should go into the Resolver site, as apt or md. And of course, there is the final question: should we do it before the release? Given the fact that our documentation does not match or dependency resolution code, it needs to be reviewed first before being merged. This should happen after 1.1.0. [1] http://maven.apache.org/resolver-archives/resolver-LATEST/ [2] http://wiki.eclipse.org/Aether Le vendredi 26 mai 2017, 16:18:02 CEST Michael Osipov a écrit : Hi folks, is there anything holding us back from MRESOLVER 1.1.0? I'd like to start the release by the end of the week and have it integrated into Maven 3.5.1. Any objections? Michael - To unsubscribe, e-mail: dev-unsubscr...@maven.apache.org For additional commands, e-mail: dev-h...@maven.apache.org - To unsubscribe, e-mail: dev-unsubscr...@maven.apache.org For additional commands, e-mail: dev-h...@maven.apache.org - To unsubscribe, e-mail: dev-unsubscr...@maven.apache.org For additional commands, e-mail: dev-h...@maven.apache.org
Re: Releasing Maven Resolver 1.1.0
are there seconders for http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/17f804d7 (aka "option 2")? Regards, Hervé Le samedi 27 mai 2017, 19:05:27 CEST Hervé BOUTEMY a écrit : > good links > yes, with this in mind, "api" is required for artifactId but should not be > added to module name: good catch, and good experience to share because that > was not so obvious > > Regards, > > Hervé > > Le samedi 27 mai 2017, 18:43:22 CEST Robert Scholte a écrit : > > There's no experience with this yet. > > > > Stephen Colebourne has written to related blogs: module naming[1] and > > modules are not artifacts[2] > > which might suggest that "api" should not be added. > > I do understand the addition of "api". And to make it worse, this is > > probably the most important artifact needing a module name. In most cases > > developers will only need this one: frameworks will handle the > > implementation part. :) > > > > Robert > > > > [1] http://blog.joda.org/2017/04/java-se-9-jpms-module-naming.html > > [2] > > http://blog.joda.org/2017/04/java-se-9-jpms-modules-are-not-artifacts.html > > > > > > On Sat, 27 May 2017 17:48:24 +0200, Hervé BOUTEMY > > > > wrote: > > > second option committed in another branch: > > > > > > option 1: > > > http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/d1724eb7 > > > > > > option 2: > > > http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/17f804d7 > > > > > > The only part that I'm not sure in option 2 is > > > org.apache.maven.resolver.api > > > > org.apache.maven.resolver: is it better to be explicit on the api or > > > implicit? > > > > > > Regards, > > > > > > Hervé > > > > > > Le samedi 27 mai 2017 17:37:03 CEST, vous avez écrit : > > >> I think I would change the following 2: > > >> > > >> org.apache.maven.resolver.connector_basic > > > >> org.apache.maven.resolver.connector.basic (in line with transport) > > >> org.apache.maven.resolver.test_util > > > >> org.apache.maven.resolver.testutil > > >> > > >> it's a matter of taste: the underscores look kind of weird, but that's > > >> probably caused because we've never used them as package names. > > >> > > >> And wondering if "api" should be changed, since this is the > > >> access-module > > >> and we don't use api in our packages: > > >> org.apache.maven.resolver.api > org.apache.maven.resolver > > >> > > >> Using a property makes it easier to configure the maven-jar-plugin, but > > >> it > > >> also makes these constants turn into variables, i.e. you could set them > > >> via system properties or cmdline args. > > >> If only we supported something like > > >> ${project.properties["AutomaticModuleName"]} > >> to > > >> mat ic-Module-Name> > > >> > > >> for the rest it's looking good. > > >> > > >> thanks > > >> Robert > > >> > > >> > > >> On Sat, 27 May 2017 17:20:15 +0200, Hervé BOUTEMY > > >> > > >> > > >> wrote: > > >> > please review and second if you think it's ok: > > >> > http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/d1724eb7 > > >> > > > >> > Regards, > > >> > > > >> > Hervé > > >> > > > >> > Le samedi 27 mai 2017, 13:24:47 CEST Hervé BOUTEMY a écrit : > > >> >> he he, Java 9 is really coming, with associated real world > > >> >> questions. > > >> >> > > >> >> Maven Artifact Resolver is one of rare Maven components that has a > > >> >> chance to > > >> >> become a collection Java 9 modules, since it was written quite > > >> > > >> recently > > >> > > >> >> and > > >> >> is pure new code as a result of Maven 3 refactoring, then does not > > >> > > >> have > > >> > > >> >> shared package names issues we have with Maven core itself. > > >> >> > > >> >> And since it is expected to be a lib for easy embedding of artifact > > >> >> resolution, making it a collection of Java 9 modules would be not > > >> > > >> only a > > >> > > >> >> great opportunity to test Java 9 modules, but it could be useful for > > >> >> people > > >> >> using it. > > >> >> > > >> >> Then I'm highly in favor of trying. > > >> >> > > >> >> Adding Automatic-Module-Name to the MANIFEST.MF looks feasible right > > >> >> now, > > >> >> without waiting much: I'm pretty sure module names will be obvious, > > >> > > >> and > > >> > > >> >> much > > >> >> better if we define them instead of waiting for automatic names. > > >> > > >> Let's > > >> > > >> >> start! MRESOLVER-26 created [1] > > >> >> > > >> >> Then there is the question of making real modules: is it feasible > > >> > > >> right > > >> > > >> >> now? > > >> >> Or would we need a delay to tweak the module descriptors? And that > > >> > > >> would > > >> > > >> >> mean that we need Java 9 to build, even if the generated bytecode > > >> >> remains > > >> >> Java 7 compatible, isn't it? Is Maven tooling ready to it? > > >> >> MRESOLVER-27 created to track the issue [2], but I'm not sure this > > >> >> is > > >> >> the > > >> >> right time to do this job, but for the next release after this 1.1.0 > > >> >> > > >>
Re: Releasing Maven Resolver 1.1.0
good links yes, with this in mind, "api" is required for artifactId but should not be added to module name: good catch, and good experience to share because that was not so obvious Regards, Hervé Le samedi 27 mai 2017, 18:43:22 CEST Robert Scholte a écrit : > There's no experience with this yet. > > Stephen Colebourne has written to related blogs: module naming[1] and > modules are not artifacts[2] > which might suggest that "api" should not be added. > I do understand the addition of "api". And to make it worse, this is > probably the most important artifact needing a module name. In most cases > developers will only need this one: frameworks will handle the > implementation part. :) > > Robert > > [1] http://blog.joda.org/2017/04/java-se-9-jpms-module-naming.html > [2] > http://blog.joda.org/2017/04/java-se-9-jpms-modules-are-not-artifacts.html > > > On Sat, 27 May 2017 17:48:24 +0200, Hervé BOUTEMY > > wrote: > > second option committed in another branch: > > > > option 1: > > http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/d1724eb7 > > > > option 2: > > http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/17f804d7 > > > > The only part that I'm not sure in option 2 is > > org.apache.maven.resolver.api > > > org.apache.maven.resolver: is it better to be explicit on the api or > > implicit? > > > > Regards, > > > > Hervé > > > > Le samedi 27 mai 2017 17:37:03 CEST, vous avez écrit : > >> I think I would change the following 2: > >> > >> org.apache.maven.resolver.connector_basic > > >> org.apache.maven.resolver.connector.basic (in line with transport) > >> org.apache.maven.resolver.test_util > org.apache.maven.resolver.testutil > >> > >> it's a matter of taste: the underscores look kind of weird, but that's > >> probably caused because we've never used them as package names. > >> > >> And wondering if "api" should be changed, since this is the > >> access-module > >> and we don't use api in our packages: > >> org.apache.maven.resolver.api > org.apache.maven.resolver > >> > >> Using a property makes it easier to configure the maven-jar-plugin, but > >> it > >> also makes these constants turn into variables, i.e. you could set them > >> via system properties or cmdline args. > >> If only we supported something like > >> ${project.properties["AutomaticModuleName"]} >> mat ic-Module-Name> > >> > >> for the rest it's looking good. > >> > >> thanks > >> Robert > >> > >> > >> On Sat, 27 May 2017 17:20:15 +0200, Hervé BOUTEMY > >> > >> > >> wrote: > >> > please review and second if you think it's ok: > >> > http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/d1724eb7 > >> > > >> > Regards, > >> > > >> > Hervé > >> > > >> > Le samedi 27 mai 2017, 13:24:47 CEST Hervé BOUTEMY a écrit : > >> >> he he, Java 9 is really coming, with associated real world questions. > >> >> > >> >> Maven Artifact Resolver is one of rare Maven components that has a > >> >> chance to > >> >> become a collection Java 9 modules, since it was written quite > >> > >> recently > >> > >> >> and > >> >> is pure new code as a result of Maven 3 refactoring, then does not > >> > >> have > >> > >> >> shared package names issues we have with Maven core itself. > >> >> > >> >> And since it is expected to be a lib for easy embedding of artifact > >> >> resolution, making it a collection of Java 9 modules would be not > >> > >> only a > >> > >> >> great opportunity to test Java 9 modules, but it could be useful for > >> >> people > >> >> using it. > >> >> > >> >> Then I'm highly in favor of trying. > >> >> > >> >> Adding Automatic-Module-Name to the MANIFEST.MF looks feasible right > >> >> now, > >> >> without waiting much: I'm pretty sure module names will be obvious, > >> > >> and > >> > >> >> much > >> >> better if we define them instead of waiting for automatic names. > >> > >> Let's > >> > >> >> start! MRESOLVER-26 created [1] > >> >> > >> >> Then there is the question of making real modules: is it feasible > >> > >> right > >> > >> >> now? > >> >> Or would we need a delay to tweak the module descriptors? And that > >> > >> would > >> > >> >> mean that we need Java 9 to build, even if the generated bytecode > >> >> remains > >> >> Java 7 compatible, isn't it? Is Maven tooling ready to it? > >> >> MRESOLVER-27 created to track the issue [2], but I'm not sure this is > >> >> the > >> >> right time to do this job, but for the next release after this 1.1.0 > >> >> > >> >> Regards, > >> >> > >> >> Hervé > >> >> > >> >> [1] https://issues.apache.org/jira/browse/MRESOLVER-26 > >> >> > >> >> [2] https://issues.apache.org/jira/browse/MRESOLVER-27 > >> >> > >> >> Le samedi 27 mai 2017, 11:58:43 CEST Robert Scholte a écrit : > >> >> > Hi, > >> >> > > >> >> > I've got a question from Remi Forax if we could add Java9 module > >> >> > descriptors to this project. > >> >> > This will be one of the first which can provide such descriptors > >> >> > >> >> since it > >> >> > >> >
Re: Releasing Maven Resolver 1.1.0
There's no experience with this yet. Stephen Colebourne has written to related blogs: module naming[1] and modules are not artifacts[2] which might suggest that "api" should not be added. I do understand the addition of "api". And to make it worse, this is probably the most important artifact needing a module name. In most cases developers will only need this one: frameworks will handle the implementation part. :) Robert [1] http://blog.joda.org/2017/04/java-se-9-jpms-module-naming.html [2] http://blog.joda.org/2017/04/java-se-9-jpms-modules-are-not-artifacts.html On Sat, 27 May 2017 17:48:24 +0200, Hervé BOUTEMY wrote: second option committed in another branch: option 1: http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/d1724eb7 option 2: http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/17f804d7 The only part that I'm not sure in option 2 is org.apache.maven.resolver.api > org.apache.maven.resolver: is it better to be explicit on the api or implicit? Regards, Hervé Le samedi 27 mai 2017 17:37:03 CEST, vous avez écrit : I think I would change the following 2: org.apache.maven.resolver.connector_basic > org.apache.maven.resolver.connector.basic (in line with transport) org.apache.maven.resolver.test_util > org.apache.maven.resolver.testutil it's a matter of taste: the underscores look kind of weird, but that's probably caused because we've never used them as package names. And wondering if "api" should be changed, since this is the access-module and we don't use api in our packages: org.apache.maven.resolver.api > org.apache.maven.resolver Using a property makes it easier to configure the maven-jar-plugin, but it also makes these constants turn into variables, i.e. you could set them via system properties or cmdline args. If only we supported something like ${project.properties["AutomaticModuleName"]} for the rest it's looking good. thanks Robert On Sat, 27 May 2017 17:20:15 +0200, Hervé BOUTEMY wrote: > please review and second if you think it's ok: > http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/d1724eb7 > > Regards, > > Hervé > > Le samedi 27 mai 2017, 13:24:47 CEST Hervé BOUTEMY a écrit : >> he he, Java 9 is really coming, with associated real world questions. >> >> Maven Artifact Resolver is one of rare Maven components that has a >> chance to >> become a collection Java 9 modules, since it was written quite recently >> and >> is pure new code as a result of Maven 3 refactoring, then does not have >> shared package names issues we have with Maven core itself. >> >> And since it is expected to be a lib for easy embedding of artifact >> resolution, making it a collection of Java 9 modules would be not only a >> great opportunity to test Java 9 modules, but it could be useful for >> people >> using it. >> >> Then I'm highly in favor of trying. >> >> Adding Automatic-Module-Name to the MANIFEST.MF looks feasible right >> now, >> without waiting much: I'm pretty sure module names will be obvious, and >> much >> better if we define them instead of waiting for automatic names. Let's >> start! MRESOLVER-26 created [1] >> >> Then there is the question of making real modules: is it feasible right >> now? >> Or would we need a delay to tweak the module descriptors? And that would >> mean that we need Java 9 to build, even if the generated bytecode >> remains >> Java 7 compatible, isn't it? Is Maven tooling ready to it? >> MRESOLVER-27 created to track the issue [2], but I'm not sure this is >> the >> right time to do this job, but for the next release after this 1.1.0 >> >> Regards, >> >> Hervé >> >> [1] https://issues.apache.org/jira/browse/MRESOLVER-26 >> >> [2] https://issues.apache.org/jira/browse/MRESOLVER-27 >> >> Le samedi 27 mai 2017, 11:58:43 CEST Robert Scholte a écrit : >> > Hi, >> > >> > I've got a question from Remi Forax if we could add Java9 module >> > descriptors to this project. >> > This will be one of the first which can provide such descriptors >> >> since it >> >> > has no required dependencies other then its own and its package >> >> structure >> >> > seems valid with the new Java9 rules. >> > >> > We haven't discussed this in general yet, but we have several projects >> > which are at the bottom of the dependency tree which should provide >> >> either >> >> > a module name or module descriptor when possible. >> > >> > Do we want to help the community by having already several libraries >> >> with >> >> > a module descriptor? >> > >> > Or we could add a Automatic-Module-Name to the MANIFEST.MF, so others >> >> can >> >> > refer to it by its official module name and we can add the descriptor >> >> once >> >> > Java9 has officially been released. (pro: doesn't require Java 9 :) ) >> > >> > Or do nothing yet... >> > >> > thanks, >> > Robert >> > >> > On Sat, 27 May 2017 11:42:32 +0200, Hervé BOUTEMY >> >> >> >> > wrote: >> > > Hi, >> > > >> > > No objection from me
Re: Releasing Maven Resolver 1.1.0
second option committed in another branch: option 1: http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/d1724eb7 option 2: http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/17f804d7 The only part that I'm not sure in option 2 is org.apache.maven.resolver.api > org.apache.maven.resolver: is it better to be explicit on the api or implicit? Regards, Hervé Le samedi 27 mai 2017 17:37:03 CEST, vous avez écrit : > I think I would change the following 2: > > org.apache.maven.resolver.connector_basic > > org.apache.maven.resolver.connector.basic (in line with transport) > org.apache.maven.resolver.test_util > org.apache.maven.resolver.testutil > > it's a matter of taste: the underscores look kind of weird, but that's > probably caused because we've never used them as package names. > > And wondering if "api" should be changed, since this is the access-module > and we don't use api in our packages: > org.apache.maven.resolver.api > org.apache.maven.resolver > > Using a property makes it easier to configure the maven-jar-plugin, but it > also makes these constants turn into variables, i.e. you could set them > via system properties or cmdline args. > If only we supported something like > ${project.properties["AutomaticModuleName"]} ic-Module-Name> > > for the rest it's looking good. > > thanks > Robert > > > On Sat, 27 May 2017 17:20:15 +0200, Hervé BOUTEMY > > wrote: > > please review and second if you think it's ok: > > http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/d1724eb7 > > > > Regards, > > > > Hervé > > > > Le samedi 27 mai 2017, 13:24:47 CEST Hervé BOUTEMY a écrit : > >> he he, Java 9 is really coming, with associated real world questions. > >> > >> Maven Artifact Resolver is one of rare Maven components that has a > >> chance to > >> become a collection Java 9 modules, since it was written quite recently > >> and > >> is pure new code as a result of Maven 3 refactoring, then does not have > >> shared package names issues we have with Maven core itself. > >> > >> And since it is expected to be a lib for easy embedding of artifact > >> resolution, making it a collection of Java 9 modules would be not only a > >> great opportunity to test Java 9 modules, but it could be useful for > >> people > >> using it. > >> > >> Then I'm highly in favor of trying. > >> > >> Adding Automatic-Module-Name to the MANIFEST.MF looks feasible right > >> now, > >> without waiting much: I'm pretty sure module names will be obvious, and > >> much > >> better if we define them instead of waiting for automatic names. Let's > >> start! MRESOLVER-26 created [1] > >> > >> Then there is the question of making real modules: is it feasible right > >> now? > >> Or would we need a delay to tweak the module descriptors? And that would > >> mean that we need Java 9 to build, even if the generated bytecode > >> remains > >> Java 7 compatible, isn't it? Is Maven tooling ready to it? > >> MRESOLVER-27 created to track the issue [2], but I'm not sure this is > >> the > >> right time to do this job, but for the next release after this 1.1.0 > >> > >> Regards, > >> > >> Hervé > >> > >> [1] https://issues.apache.org/jira/browse/MRESOLVER-26 > >> > >> [2] https://issues.apache.org/jira/browse/MRESOLVER-27 > >> > >> Le samedi 27 mai 2017, 11:58:43 CEST Robert Scholte a écrit : > >> > Hi, > >> > > >> > I've got a question from Remi Forax if we could add Java9 module > >> > descriptors to this project. > >> > This will be one of the first which can provide such descriptors > >> > >> since it > >> > >> > has no required dependencies other then its own and its package > >> > >> structure > >> > >> > seems valid with the new Java9 rules. > >> > > >> > We haven't discussed this in general yet, but we have several projects > >> > which are at the bottom of the dependency tree which should provide > >> > >> either > >> > >> > a module name or module descriptor when possible. > >> > > >> > Do we want to help the community by having already several libraries > >> > >> with > >> > >> > a module descriptor? > >> > > >> > Or we could add a Automatic-Module-Name to the MANIFEST.MF, so others > >> > >> can > >> > >> > refer to it by its official module name and we can add the descriptor > >> > >> once > >> > >> > Java9 has officially been released. (pro: doesn't require Java 9 :) ) > >> > > >> > Or do nothing yet... > >> > > >> > thanks, > >> > Robert > >> > > >> > On Sat, 27 May 2017 11:42:32 +0200, Hervé BOUTEMY > >> > >> > >> > >> > wrote: > >> > > Hi, > >> > > > >> > > No objection from me, thanks for keeping the ball rolling. > >> > > > >> > > I tried to improve documentation by adding some useful links to > >> > >> other > >> > >> > > related > >> > > components [1]: I think the current state is better and ok for a > >> > > release. > >> > > > >> > > One key question now is about Aether wiki content [2]: should we > >> > >> copy > >> > >> > > it? In a > >> > > wiki or
Re: Releasing Maven Resolver 1.1.0
I think I would change the following 2: org.apache.maven.resolver.connector_basic > org.apache.maven.resolver.connector.basic (in line with transport) org.apache.maven.resolver.test_util > org.apache.maven.resolver.testutil it's a matter of taste: the underscores look kind of weird, but that's probably caused because we've never used them as package names. And wondering if "api" should be changed, since this is the access-module and we don't use api in our packages: org.apache.maven.resolver.api > org.apache.maven.resolver Using a property makes it easier to configure the maven-jar-plugin, but it also makes these constants turn into variables, i.e. you could set them via system properties or cmdline args. If only we supported something like ${project.properties["AutomaticModuleName"]} for the rest it's looking good. thanks Robert On Sat, 27 May 2017 17:20:15 +0200, Hervé BOUTEMY wrote: please review and second if you think it's ok: http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/d1724eb7 Regards, Hervé Le samedi 27 mai 2017, 13:24:47 CEST Hervé BOUTEMY a écrit : he he, Java 9 is really coming, with associated real world questions. Maven Artifact Resolver is one of rare Maven components that has a chance to become a collection Java 9 modules, since it was written quite recently and is pure new code as a result of Maven 3 refactoring, then does not have shared package names issues we have with Maven core itself. And since it is expected to be a lib for easy embedding of artifact resolution, making it a collection of Java 9 modules would be not only a great opportunity to test Java 9 modules, but it could be useful for people using it. Then I'm highly in favor of trying. Adding Automatic-Module-Name to the MANIFEST.MF looks feasible right now, without waiting much: I'm pretty sure module names will be obvious, and much better if we define them instead of waiting for automatic names. Let's start! MRESOLVER-26 created [1] Then there is the question of making real modules: is it feasible right now? Or would we need a delay to tweak the module descriptors? And that would mean that we need Java 9 to build, even if the generated bytecode remains Java 7 compatible, isn't it? Is Maven tooling ready to it? MRESOLVER-27 created to track the issue [2], but I'm not sure this is the right time to do this job, but for the next release after this 1.1.0 Regards, Hervé [1] https://issues.apache.org/jira/browse/MRESOLVER-26 [2] https://issues.apache.org/jira/browse/MRESOLVER-27 Le samedi 27 mai 2017, 11:58:43 CEST Robert Scholte a écrit : > Hi, > > I've got a question from Remi Forax if we could add Java9 module > descriptors to this project. > This will be one of the first which can provide such descriptors since it > has no required dependencies other then its own and its package structure > seems valid with the new Java9 rules. > > We haven't discussed this in general yet, but we have several projects > which are at the bottom of the dependency tree which should provide either > a module name or module descriptor when possible. > > Do we want to help the community by having already several libraries with > a module descriptor? > > Or we could add a Automatic-Module-Name to the MANIFEST.MF, so others can > refer to it by its official module name and we can add the descriptor once > Java9 has officially been released. (pro: doesn't require Java 9 :) ) > > Or do nothing yet... > > thanks, > Robert > > On Sat, 27 May 2017 11:42:32 +0200, Hervé BOUTEMY > > wrote: > > Hi, > > > > No objection from me, thanks for keeping the ball rolling. > > > > I tried to improve documentation by adding some useful links to other > > related > > components [1]: I think the current state is better and ok for a > > release. > > > > One key question now is about Aether wiki content [2]: should we copy > > it? In a > > wiki or in components sources? > > I suppose wiki source format is supported by Doxia, then it could be > > imported > > quite easily in sources. > > > > And of course, there is the final question: should we do it before the > > release? > > > > Regards, > > > > Hervé > > > > [1] http://maven.apache.org/resolver-archives/resolver-LATEST/ > > > > [2] http://wiki.eclipse.org/Aether > > > > Le vendredi 26 mai 2017, 16:18:02 CEST Michael Osipov a écrit : > >> Hi folks, > >> > >> is there anything holding us back from MRESOLVER 1.1.0? > >> I'd like to start the release by the end of the week and have it > >> integrated into Maven 3.5.1. > >> > >> Any objections? > >> > >> Michael > >> > >> - > >> To unsubscribe, e-mail: dev-unsubscr...@maven.apache.org > >> For additional commands, e-mail: dev-h...@maven.apache.org > > > > - > > To unsubscribe, e-mail: dev-unsubscr...@maven.apache.org > > For ad
Re: Releasing Maven Resolver 1.1.0
please review and second if you think it's ok: http://git-wip-us.apache.org/repos/asf/maven-resolver/commit/d1724eb7 Regards, Hervé Le samedi 27 mai 2017, 13:24:47 CEST Hervé BOUTEMY a écrit : > he he, Java 9 is really coming, with associated real world questions. > > Maven Artifact Resolver is one of rare Maven components that has a chance to > become a collection Java 9 modules, since it was written quite recently and > is pure new code as a result of Maven 3 refactoring, then does not have > shared package names issues we have with Maven core itself. > > And since it is expected to be a lib for easy embedding of artifact > resolution, making it a collection of Java 9 modules would be not only a > great opportunity to test Java 9 modules, but it could be useful for people > using it. > > Then I'm highly in favor of trying. > > Adding Automatic-Module-Name to the MANIFEST.MF looks feasible right now, > without waiting much: I'm pretty sure module names will be obvious, and much > better if we define them instead of waiting for automatic names. Let's > start! MRESOLVER-26 created [1] > > Then there is the question of making real modules: is it feasible right now? > Or would we need a delay to tweak the module descriptors? And that would > mean that we need Java 9 to build, even if the generated bytecode remains > Java 7 compatible, isn't it? Is Maven tooling ready to it? > MRESOLVER-27 created to track the issue [2], but I'm not sure this is the > right time to do this job, but for the next release after this 1.1.0 > > Regards, > > Hervé > > [1] https://issues.apache.org/jira/browse/MRESOLVER-26 > > [2] https://issues.apache.org/jira/browse/MRESOLVER-27 > > Le samedi 27 mai 2017, 11:58:43 CEST Robert Scholte a écrit : > > Hi, > > > > I've got a question from Remi Forax if we could add Java9 module > > descriptors to this project. > > This will be one of the first which can provide such descriptors since it > > has no required dependencies other then its own and its package structure > > seems valid with the new Java9 rules. > > > > We haven't discussed this in general yet, but we have several projects > > which are at the bottom of the dependency tree which should provide either > > a module name or module descriptor when possible. > > > > Do we want to help the community by having already several libraries with > > a module descriptor? > > > > Or we could add a Automatic-Module-Name to the MANIFEST.MF, so others can > > refer to it by its official module name and we can add the descriptor once > > Java9 has officially been released. (pro: doesn't require Java 9 :) ) > > > > Or do nothing yet... > > > > thanks, > > Robert > > > > On Sat, 27 May 2017 11:42:32 +0200, Hervé BOUTEMY > > > > wrote: > > > Hi, > > > > > > No objection from me, thanks for keeping the ball rolling. > > > > > > I tried to improve documentation by adding some useful links to other > > > related > > > components [1]: I think the current state is better and ok for a > > > release. > > > > > > One key question now is about Aether wiki content [2]: should we copy > > > it? In a > > > wiki or in components sources? > > > I suppose wiki source format is supported by Doxia, then it could be > > > imported > > > quite easily in sources. > > > > > > And of course, there is the final question: should we do it before the > > > release? > > > > > > Regards, > > > > > > Hervé > > > > > > [1] http://maven.apache.org/resolver-archives/resolver-LATEST/ > > > > > > [2] http://wiki.eclipse.org/Aether > > > > > > Le vendredi 26 mai 2017, 16:18:02 CEST Michael Osipov a écrit : > > >> Hi folks, > > >> > > >> is there anything holding us back from MRESOLVER 1.1.0? > > >> I'd like to start the release by the end of the week and have it > > >> integrated into Maven 3.5.1. > > >> > > >> Any objections? > > >> > > >> Michael > > >> > > >> - > > >> To unsubscribe, e-mail: dev-unsubscr...@maven.apache.org > > >> For additional commands, e-mail: dev-h...@maven.apache.org > > > > > > - > > > To unsubscribe, e-mail: dev-unsubscr...@maven.apache.org > > > For additional commands, e-mail: dev-h...@maven.apache.org > > > > - > > To unsubscribe, e-mail: dev-unsubscr...@maven.apache.org > > For additional commands, e-mail: dev-h...@maven.apache.org > > - > To unsubscribe, e-mail: dev-unsubscr...@maven.apache.org > For additional commands, e-mail: dev-h...@maven.apache.org - To unsubscribe, e-mail: dev-unsubscr...@maven.apache.org For additional commands, e-mail: dev-h...@maven.apache.org
Re: Releasing Maven Resolver 1.1.0
On Sat, 27 May 2017 13:24:47 +0200, Hervé BOUTEMY wrote: he he, Java 9 is really coming, with associated real world questions. Maven Artifact Resolver is one of rare Maven components that has a chance to become a collection Java 9 modules, since it was written quite recently and is pure new code as a result of Maven 3 refactoring, then does not have shared package names issues we have with Maven core itself. And since it is expected to be a lib for easy embedding of artifact resolution, making it a collection of Java 9 modules would be not only a great opportunity to test Java 9 modules, but it could be useful for people using it. Then I'm highly in favor of trying. Adding Automatic-Module-Name to the MANIFEST.MF looks feasible right now, without waiting much: I'm pretty sure module names will be obvious, and much better if we define them instead of waiting for automatic names. Let's start! MRESOLVER-26 created [1] This was indeed the approach I had in mind right now :) Then there is the question of making real modules: is it feasible right now? Or would we need a delay to tweak the module descriptors? And that would mean that we need Java 9 to build, even if the generated bytecode remains Java 7 compatible, isn't it? Is Maven tooling ready to it? The module-info.java MUST be compiled with Java 9 (source/target/release=9), while the rest of the classes can stay at a lower version. This implies 2 separate executions as described on the maven-compler-plugin page[1]. Let's start doing this after the official release of J9 thanks, Robert [1] http://maven.apache.org/plugins/maven-compiler-plugin/examples/module-info.html MRESOLVER-27 created to track the issue [2], but I'm not sure this is the right time to do this job, but for the next release after this 1.1.0 Regards, Hervé [1] https://issues.apache.org/jira/browse/MRESOLVER-26 [2] https://issues.apache.org/jira/browse/MRESOLVER-27 Le samedi 27 mai 2017, 11:58:43 CEST Robert Scholte a écrit : Hi, I've got a question from Remi Forax if we could add Java9 module descriptors to this project. This will be one of the first which can provide such descriptors since it has no required dependencies other then its own and its package structure seems valid with the new Java9 rules. We haven't discussed this in general yet, but we have several projects which are at the bottom of the dependency tree which should provide either a module name or module descriptor when possible. Do we want to help the community by having already several libraries with a module descriptor? Or we could add a Automatic-Module-Name to the MANIFEST.MF, so others can refer to it by its official module name and we can add the descriptor once Java9 has officially been released. (pro: doesn't require Java 9 :) ) Or do nothing yet... thanks, Robert On Sat, 27 May 2017 11:42:32 +0200, Hervé BOUTEMY wrote: > Hi, > > No objection from me, thanks for keeping the ball rolling. > > I tried to improve documentation by adding some useful links to other > related > components [1]: I think the current state is better and ok for a release. > > One key question now is about Aether wiki content [2]: should we copy > it? In a > wiki or in components sources? > I suppose wiki source format is supported by Doxia, then it could be > imported > quite easily in sources. > > And of course, there is the final question: should we do it before the > release? > > Regards, > > Hervé > > [1] http://maven.apache.org/resolver-archives/resolver-LATEST/ > > [2] http://wiki.eclipse.org/Aether > > Le vendredi 26 mai 2017, 16:18:02 CEST Michael Osipov a écrit : >> Hi folks, >> >> is there anything holding us back from MRESOLVER 1.1.0? >> I'd like to start the release by the end of the week and have it >> integrated into Maven 3.5.1. >> >> Any objections? >> >> Michael >> >> - >> To unsubscribe, e-mail: dev-unsubscr...@maven.apache.org >> For additional commands, e-mail: dev-h...@maven.apache.org > > - > To unsubscribe, e-mail: dev-unsubscr...@maven.apache.org > For additional commands, e-mail: dev-h...@maven.apache.org - To unsubscribe, e-mail: dev-unsubscr...@maven.apache.org For additional commands, e-mail: dev-h...@maven.apache.org - To unsubscribe, e-mail: dev-unsubscr...@maven.apache.org For additional commands, e-mail: dev-h...@maven.apache.org - To unsubscribe, e-mail: dev-unsubscr...@maven.apache.org For additional commands, e-mail: dev-h...@maven.apache.org
Re: Releasing Maven Resolver 1.1.0
he he, Java 9 is really coming, with associated real world questions. Maven Artifact Resolver is one of rare Maven components that has a chance to become a collection Java 9 modules, since it was written quite recently and is pure new code as a result of Maven 3 refactoring, then does not have shared package names issues we have with Maven core itself. And since it is expected to be a lib for easy embedding of artifact resolution, making it a collection of Java 9 modules would be not only a great opportunity to test Java 9 modules, but it could be useful for people using it. Then I'm highly in favor of trying. Adding Automatic-Module-Name to the MANIFEST.MF looks feasible right now, without waiting much: I'm pretty sure module names will be obvious, and much better if we define them instead of waiting for automatic names. Let's start! MRESOLVER-26 created [1] Then there is the question of making real modules: is it feasible right now? Or would we need a delay to tweak the module descriptors? And that would mean that we need Java 9 to build, even if the generated bytecode remains Java 7 compatible, isn't it? Is Maven tooling ready to it? MRESOLVER-27 created to track the issue [2], but I'm not sure this is the right time to do this job, but for the next release after this 1.1.0 Regards, Hervé [1] https://issues.apache.org/jira/browse/MRESOLVER-26 [2] https://issues.apache.org/jira/browse/MRESOLVER-27 Le samedi 27 mai 2017, 11:58:43 CEST Robert Scholte a écrit : > Hi, > > I've got a question from Remi Forax if we could add Java9 module > descriptors to this project. > This will be one of the first which can provide such descriptors since it > has no required dependencies other then its own and its package structure > seems valid with the new Java9 rules. > > We haven't discussed this in general yet, but we have several projects > which are at the bottom of the dependency tree which should provide either > a module name or module descriptor when possible. > > Do we want to help the community by having already several libraries with > a module descriptor? > > Or we could add a Automatic-Module-Name to the MANIFEST.MF, so others can > refer to it by its official module name and we can add the descriptor once > Java9 has officially been released. (pro: doesn't require Java 9 :) ) > > Or do nothing yet... > > thanks, > Robert > > On Sat, 27 May 2017 11:42:32 +0200, Hervé BOUTEMY > > wrote: > > Hi, > > > > No objection from me, thanks for keeping the ball rolling. > > > > I tried to improve documentation by adding some useful links to other > > related > > components [1]: I think the current state is better and ok for a release. > > > > One key question now is about Aether wiki content [2]: should we copy > > it? In a > > wiki or in components sources? > > I suppose wiki source format is supported by Doxia, then it could be > > imported > > quite easily in sources. > > > > And of course, there is the final question: should we do it before the > > release? > > > > Regards, > > > > Hervé > > > > [1] http://maven.apache.org/resolver-archives/resolver-LATEST/ > > > > [2] http://wiki.eclipse.org/Aether > > > > Le vendredi 26 mai 2017, 16:18:02 CEST Michael Osipov a écrit : > >> Hi folks, > >> > >> is there anything holding us back from MRESOLVER 1.1.0? > >> I'd like to start the release by the end of the week and have it > >> integrated into Maven 3.5.1. > >> > >> Any objections? > >> > >> Michael > >> > >> - > >> To unsubscribe, e-mail: dev-unsubscr...@maven.apache.org > >> For additional commands, e-mail: dev-h...@maven.apache.org > > > > - > > To unsubscribe, e-mail: dev-unsubscr...@maven.apache.org > > For additional commands, e-mail: dev-h...@maven.apache.org > > - > To unsubscribe, e-mail: dev-unsubscr...@maven.apache.org > For additional commands, e-mail: dev-h...@maven.apache.org - To unsubscribe, e-mail: dev-unsubscr...@maven.apache.org For additional commands, e-mail: dev-h...@maven.apache.org
Re: Releasing Maven Resolver 1.1.0
Hi, I've got a question from Remi Forax if we could add Java9 module descriptors to this project. This will be one of the first which can provide such descriptors since it has no required dependencies other then its own and its package structure seems valid with the new Java9 rules. We haven't discussed this in general yet, but we have several projects which are at the bottom of the dependency tree which should provide either a module name or module descriptor when possible. Do we want to help the community by having already several libraries with a module descriptor? Or we could add a Automatic-Module-Name to the MANIFEST.MF, so others can refer to it by its official module name and we can add the descriptor once Java9 has officially been released. (pro: doesn't require Java 9 :) ) Or do nothing yet... thanks, Robert On Sat, 27 May 2017 11:42:32 +0200, Hervé BOUTEMY wrote: Hi, No objection from me, thanks for keeping the ball rolling. I tried to improve documentation by adding some useful links to other related components [1]: I think the current state is better and ok for a release. One key question now is about Aether wiki content [2]: should we copy it? In a wiki or in components sources? I suppose wiki source format is supported by Doxia, then it could be imported quite easily in sources. And of course, there is the final question: should we do it before the release? Regards, Hervé [1] http://maven.apache.org/resolver-archives/resolver-LATEST/ [2] http://wiki.eclipse.org/Aether Le vendredi 26 mai 2017, 16:18:02 CEST Michael Osipov a écrit : Hi folks, is there anything holding us back from MRESOLVER 1.1.0? I'd like to start the release by the end of the week and have it integrated into Maven 3.5.1. Any objections? Michael - To unsubscribe, e-mail: dev-unsubscr...@maven.apache.org For additional commands, e-mail: dev-h...@maven.apache.org - To unsubscribe, e-mail: dev-unsubscr...@maven.apache.org For additional commands, e-mail: dev-h...@maven.apache.org - To unsubscribe, e-mail: dev-unsubscr...@maven.apache.org For additional commands, e-mail: dev-h...@maven.apache.org
Re: Releasing Maven Resolver 1.1.0
Hi, No objection from me, thanks for keeping the ball rolling. I tried to improve documentation by adding some useful links to other related components [1]: I think the current state is better and ok for a release. One key question now is about Aether wiki content [2]: should we copy it? In a wiki or in components sources? I suppose wiki source format is supported by Doxia, then it could be imported quite easily in sources. And of course, there is the final question: should we do it before the release? Regards, Hervé [1] http://maven.apache.org/resolver-archives/resolver-LATEST/ [2] http://wiki.eclipse.org/Aether Le vendredi 26 mai 2017, 16:18:02 CEST Michael Osipov a écrit : > Hi folks, > > is there anything holding us back from MRESOLVER 1.1.0? > I'd like to start the release by the end of the week and have it > integrated into Maven 3.5.1. > > Any objections? > > Michael > > - > To unsubscribe, e-mail: dev-unsubscr...@maven.apache.org > For additional commands, e-mail: dev-h...@maven.apache.org - To unsubscribe, e-mail: dev-unsubscr...@maven.apache.org For additional commands, e-mail: dev-h...@maven.apache.org