Maven 2 & 3

site reporting causes release:perform to fail

Details

  • Type: Bug Bug
  • Status: Open Open
  • Priority: Major Major
  • Resolution: Unresolved
  • Affects Version/s: None
  • Fix Version/s: Issues to be reviewed for 3.x
  • Component/s: None
  • Labels:
    None
  • Number of attachments :
    2

Description

It seems that if mxing aggregates and non-aggregates of the same report that it causes release:perform to fail! This definitely happens with javadoc .
Either removing the aggregate or non-aggregate version seems to not cause a release:perform failure.

The specific command lines I am using:
mvn -B -Dresume=false -Dmaven.test.skip release:prepare
mvn -B -Dmaven.test.skip release:perform

I can also get this error when doing mvn site:site if I have not done a mvn install since the version was updated.

Activity

Ascending order - Click to sort in descending order
Hide
Permalink
Paul Sundling added a comment -

Forgot to mention, zip file contains code that reproduces issue with aggregate reports causing release failure.

Show
Paul Sundling added a comment - Forgot to mention, zip file contains code that reproduces issue with aggregate reports causing release failure.
Brett Porter made changes -
Field Original Value New Value
Fix Version/s 2.0.x [ 13141 ]
Hide
Permalink
Paul Sundling added a comment -

It's been a year. Has anyone looked into the bug since then?

Show
Paul Sundling added a comment - It's been a year. Has anyone looked into the bug since then?
Hide
Permalink
Paul Sundling added a comment -

This branch was created to test releases. For more information see http://jira.codehaus.org/browse/MNG-3336.

One piece of the puzzle is that I have to run install before the release:prepare, see http://jira.codehaus.org/browse/MJAVADOC-116 for more details. Also tests can't be skipped with release plugin, so I've removed the flag. So here are the updated command lines to test this branch.

mvn -B -Dresume=false install release:prepare
mvn -B release:perform

This branch uses PUBLISH_DIR as a variable for where to publish stuff. So set this environment variable.

Other unrelated issues encountered:

PMD 2.4 bug where it complains about annotation if you don't set targetJdk to 1.5 (1.6 or anything else blows up during site, but works for pmd:pmd!).

Perforce/SCM plugin issues (must have been transitively upgraded or something): now instead of whatever mechanism, perforce expects P4PASSWD to be your clear text password. So probably perforce issue mostly, just a weird coincidence.

If encounter problems with permission denied, try it on linux, which is a copy on write operating system. Windows will lock files causing permissions problems if another process has changed it or is used as a working directory.

This attached zip file has a configuration that worked. It includes a lot more reporting, but given how many attempts it took to get a working config, I'm not going to mess with the magic sauce.

Show
Paul Sundling added a comment - This branch was created to test releases. For more information see http://jira.codehaus.org/browse/MNG-3336 . One piece of the puzzle is that I have to run install before the release:prepare, see http://jira.codehaus.org/browse/MJAVADOC-116 for more details. Also tests can't be skipped with release plugin, so I've removed the flag. So here are the updated command lines to test this branch. mvn -B -Dresume=false install release:prepare mvn -B release:perform This branch uses PUBLISH_DIR as a variable for where to publish stuff. So set this environment variable. Other unrelated issues encountered: PMD 2.4 bug where it complains about annotation if you don't set targetJdk to 1.5 (1.6 or anything else blows up during site, but works for pmd:pmd!). Perforce/SCM plugin issues (must have been transitively upgraded or something): now instead of whatever mechanism, perforce expects P4PASSWD to be your clear text password. So probably perforce issue mostly, just a weird coincidence. If encounter problems with permission denied, try it on linux, which is a copy on write operating system. Windows will lock files causing permissions problems if another process has changed it or is used as a working directory. This attached zip file has a configuration that worked. It includes a lot more reporting, but given how many attempts it took to get a working config, I'm not going to mess with the magic sauce.
Hide
Permalink
Paul Sundling added a comment -

This is the zip with working config.

Show
Paul Sundling added a comment - This is the zip with working config.
Paul Sundling made changes -
Attachment minimal.zip [ 39473 ]
Hide
Permalink
Benjamin Bentmann added a comment -

Also tests can't be skipped with release plugin, so I've removed the flag.

The Release Plugin doesn't automatically propagate all system properties to the forked build, so one needs to use the arguments parameter.

Show
Benjamin Bentmann added a comment - Also tests can't be skipped with release plugin, so I've removed the flag. The Release Plugin doesn't automatically propagate all system properties to the forked build, so one needs to use the arguments parameter.
Hide
Permalink
Paul Sundling added a comment -

Forgot to mention one thing. The working config has only aggregate javadocs and although it has only aggregate jxr reports, in my real production config I was able to mix aggregate and non-aggregate jxr reports. I didn't try mixing javadoc aggregate and non-aggregate, as this was good enough. Or maybe I did. It took lots of experimenting to get something functional and by the end I was so frustrated I didn't know which way was up, much less remember all the various 50 permutations or so.

Show
Paul Sundling added a comment - Forgot to mention one thing. The working config has only aggregate javadocs and although it has only aggregate jxr reports, in my real production config I was able to mix aggregate and non-aggregate jxr reports. I didn't try mixing javadoc aggregate and non-aggregate, as this was good enough. Or maybe I did. It took lots of experimenting to get something functional and by the end I was so frustrated I didn't know which way was up, much less remember all the various 50 permutations or so.
Brett Porter made changes -
Fix Version/s Issues to be reviewed for 3.x [ 13145 ]
Fix Version/s 2.2.x (to be reviewed) [ 13141 ]

People

  • Assignee:
    Unassigned
    Reporter:
    Paul Sundling
Vote (0)
Watch (1)

Dates

  • Created:
    Updated: