[ 
https://issues.apache.org/jira/browse/MSHARED-1285?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17792200#comment-17792200
 ] 

ASF GitHub Bot commented on MSHARED-1285:
-----------------------------------------

lalmeras commented on PR #77:
URL: https://github.com/apache/maven-filtering/pull/77#issuecomment-1836797752

   I update my branch 
https://github.com/lalmeras/maven-filtering/commits/MSHARED-1285 with the 
following change :
   * rebased on origin/master with the refactored tests
   * I added a test-case demonstrating the usecase described by @laeubi : if 
target files are removed, an incremental build fails to regenerate missing 
resources
   
   > Another case is that you modify the target file (but not the source) then 
the mojo will never update that file as long as one do not modify the source 
file.
   
   @laeubi I'm not sure to understand what is the current / expected behavior 
for this use-case. Can you give a more precise description ?




> DefaultMavenResourcesFiltering uses BuildContext in a way that fails sometimes
> ------------------------------------------------------------------------------
>
>                 Key: MSHARED-1285
>                 URL: https://issues.apache.org/jira/browse/MSHARED-1285
>             Project: Maven Shared Components
>          Issue Type: Bug
>            Reporter: Christoph Läubrich
>            Priority: Major
>
> The maven resources plugin uses 
> [https://github.com/apache/maven-filtering/blob/master/src/main/java/org/apache/maven/shared/filtering/DefaultMavenResourcesFiltering.java]
>  to copy resources, but that component has some subtile flaws reported here:
> [https://github.com/eclipse-m2e/m2e-core/discussions/1468]
> The problematic part is the usage of BuildContext#newScanner that already 
> mentions in the javadoc that passing ignoreDelta to neScanner might not 
> reveal all required items +*for copy-resources*+ form A -> B and instead 
> BuildContext#isUpTodate should be used.
> Just from a quick look I assume that part of code actually wants to use 
> something like this:
> {code:java}
> DirectoryScanner ds = new DirectoryScanner() {
>   @Override
>   protected boolean isSelected(String name, File file) {
>     if (file.isFile() && buildContext.isUptodate(getTargetFile(file), file)) 
> { 
>      return false;
>     }
>     return true;
>   }
> };
> ds.setBasedir(basedir); {code}
> That way all the code that currently checks for if output directory existed 
> before can also be removed what is another issue because it leads to a full 
> resources copy even if source+target are already even if a single class file 
> is changed while the idea seems more that if the output was not there it 
> should not try to be incremental!



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

Reply via email to