On 1 February 2017 at 10:34, Karl Heinz Marbaise <[email protected]> wrote:
> Hi Stephen, > > On 01/02/17 11:22, Stephen Connolly wrote: > >> >> >> On 1 February 2017 at 10:15, Karl Heinz Marbaise <[email protected] >> <mailto:[email protected]>> wrote: >> >> Hi, >> >> On 01/02/17 10:50, Stephen Connolly wrote: >> >> But doesn't the shade plugin want to do exactly that... namely >> attach >> the shaded jar *to replace* the jar attached by the jar plugin? >> >> >> Yes but if the maven-source-plugin does the same and attach a source >> package which means you have two artifacts being attached with the >> same name... >> >> This can in the end result in build failures during a release build >> where those artifacts are being transfered to repo managaer which >> than blocks for the second artifact which is already been there... >> >> The install plugin will simply install the same artifact a second >> time and overwrites on HD so no problem...but deploy plugin will try >> to upload the same artifact twice... >> >> >> Yes so the bug was where the attach artifacts with the same coordinates >> as an already attached artifact *appends* rather than updating >> >> If it was an update rather than an append, then there would only be one >> artifact for install and deploy to install and deploy. >> >> This may mean that we need a new API method to update an attached >> artifact and issue a warning for the use of the non-update variant, but >> to allow existing plugins to behave the intended way (before the change >> that clobbered the behaviour and made attach append always) we would >> apply the update. >> > > Yes... > > So this would be the first step to produce a WARNING and let people > reconsider their builds or fix their plugins...and of course enhance the > API and with future versions of Maven (which? 3.6.0 for example? should be > discussed) we should fail the build ... I disagree with failing the build... We should fail the build if the plugin was compiled against the API that provides the "attachOrUpdate" method... but if the plugin was compiled against the APIs that only provided the "attach" method then we should probably not even warn and just assume they meant "attachOrUpdate" > > Kind regards > Karl Heinz Marbaise > > > >> >> >> I've seen that also in different combinations like people added >> several plugins building the same artifact with different plugins >> and weren't aware that there is something wrong and later the >> release build failed... >> >> >> Kind regards >> Karl Heinz Marbaise >> >> >> >> >> On 1 February 2017 at 08:51, Karl Heinz Marbaise >> <[email protected] <mailto:[email protected]> >> <mailto:[email protected] <mailto:[email protected]>>> wrote: >> >> Hi, >> >> fixing this issue will produce a WARNING in cases where >> different >> plugins attach the same artifact multiple times which gives >> a hint >> on a wrong build configuration... >> >> Examples: maven-sources-plugin attaches a jar and another >> plugin >> will do the same thing...which results in complains that >> attached >> jars are missing...But the reason is that plugins are doing >> it wrong >> and do not realize they are doing something wrong...or the >> user does... >> >> But I don't see a problem to move this to 3.5.1 ...Moved so to >> 3.5.1...and reopened it.... >> >> Kind regards >> Karl Heinz Marbaise >> >> >> On 01/02/17 08:58, Anders Hammar wrote: >> >> Didn't we agree on v3.5.0 to be a drop-in replacement for >> v3.3.9? IMO >> fixing MNG-5868 wouldn't fit in that. >> >> I'm sorry to say, but I think we're heading back to where >> we >> were before >> the reset. >> >> /Anders >> >> On Wed, Feb 1, 2017 at 8:42 AM, Hervé BOUTEMY >> <[email protected] <mailto:[email protected]> >> <mailto:[email protected] <mailto:[email protected]>>> >> >> wrote: >> >> https://issues.apache.org/jira/browse/MNG-5868 >> <https://issues.apache.org/jira/browse/MNG-5868> >> >> <https://issues.apache.org/jira/browse/MNG-5868 >> <https://issues.apache.org/jira/browse/MNG-5868>> >> Adding serval times the same artifact via >> MavenProjectHelper >> (attachArtifact) >> does not produce a failure >> >> by reading the Jira entry, I can't understand what >> has been >> done and what >> is >> the effective impact: IIUC, Maven core becomes more >> picky, >> expectedly to >> help >> users discover unexpected situations by failing >> instead of >> silently doing >> something that seems odd. But how many plugins are >> affected? >> What will >> users >> get as a result? Are there some plugins versions >> that are to >> be upgraded? >> >> not clear from the Jira issue >> >> I need more explanations in Jira before saying if >> this is >> the right thing >> to >> do, whatever Maven version we are targetting >> >> Regards, >> >> Hervé >> >> Le mercredi 1 février 2017, 00:05:01 CET Christian >> Schulte a >> écrit : >> >> Hi, >> >> I'd like to make MNG-5868 FIX-3.5.0. There have >> been >> plugin issues >> solved by this. I know Karl-Heinz worked on some >> of >> those plugin issues. >> If this does not get released, those plugin >> issues may >> need to be >> re-opened. Anyone second FIX-3.5.0? >> >> Regards, >> >>
