Thanks for the info guys, just trying to understand what's happening here. I've just looked at the ant task [1] and the original bug [2] for adding apt to ant. Their discussion of the 'preprocessdir' parameter in the bug comments has clarified the somewhat confusing distinction between -d and -s in the JDK documentation [3]. To emphasize:
-d = directory for processor generated *resource* files (and compiled class files, but we never generate any since we always specify -nocompile) -s = directory for processor generated *java* files I assume that these locations correlate to CLASS_TREE and SOURCE_TREE respectively in com.sun.mirror.apt.Filer.Location. All the current tests generate text files in the -s SOURCE_TREE, which now seems wrong as they should be placed in the -d CLASS_TREE. So in light of all this, I think we need to: 1) change the plugin's outputDirectory parameter to set -d instead of -s 2) add a sourceOutputDirectory parameter to explicit set -s 3) add a test for sourceOutputDirectory Does this match everyone's understanding of apt? Cheers, Mark [1] http://ant.apache.org/manual/CoreTasks/apt.html [2] https://issues.apache.org/bugzilla/show_bug.cgi?id=29978 [3] http://java.sun.com/j2se/1.5.0/docs/tooldocs/share/apt.html 2010/1/11 Peter Lynch <[email protected]>: > Mark, here is what I know so far about this problem. > > - only reproduced in Java 1.6, but not always > - happened on Mac OSX 10.5 and WinXP cygwin from two different people > - appears inconsistently > - when it appears, simply running the existing integration tests for the apt > plugin will reproduce the problem ( 16 of 18 tests will fail!) > - patch to fix it mimics closely what Ant does in its APT task. > - apt plugin code as is does not mimic what Ant task does. > - there are no related open issues to Ant apt task > - Ant apt task has been around for 5 years > - when current patch is applied, new problems do not appear. > > Please understand I will not provide a unique integration test for this bug > since 16 of the 18 existing integration tests will fail if the bug appears > anyways. > > Given the above, perhaps consider applying the patch and taking our chances? > > -Peter > > On Mon, Jan 11, 2010 at 10:42 AM, Mark Hobson <[email protected]> wrote: >> >> Ryan, what environment were you using in that stack trace you posted? >> (OS, JVM, Maven.) >> >> Mark >> >> 2010/1/6 Mark Hobson <[email protected]>: >> > How strange, thanks for retrying Peter. I'd rather not apply a patch >> > without it demonstrably fixing something as it just confuses the code. >> > >> > Cheers, >> > >> > Mark >> > >> > 2010/1/6 Peter Lynch <[email protected]>: >> >> I just went back and tested 1.0-alpha-3 again on my project and for >> >> some >> >> reason it works fine without any changes. This is really weird. >> >> Although I >> >> think the patch is harmless, I'll mark the issue as Closed, can't >> >> reproduce. >> >> >> >> Thanks. >> >> >> >> On Tue, Jan 5, 2010 at 4:31 PM, Mark Hobson <[email protected]> >> >> wrote: >> >>> >> >>> Hi Peter, >> >>> >> >>> Apologies for the late reply. I'm still unsure as to what problem >> >>> you're trying to fix here. Can you supply details on how to reproduce >> >>> it? I've been using this plugin successfully for a while now. >> >>> >> >>> Cheers, >> >>> >> >>> Mark >> >>> >> >>> 2009/12/17 Peter Lynch <[email protected]>: >> >>> > Hi Mark, thanks for taking interest in my troubles. >> >>> > >> >>> > See my updates to the issue with a new patch. If you run the tests >> >>> > with >> >>> > the >> >>> > new patch applied then the changes get suitably exercised by the >> >>> > ITs. >> >>> > Preferably you can run it on something other than OSX to verify. >> >>> > >> >>> > Let me know if I can be of more help. >> >>> > >> >>> > -Peter >> >>> > >> >>> > On Thu, Dec 17, 2009 at 5:09 AM, Mark Hobson <[email protected]> >> >>> > wrote: >> >>> >> >> >>> >> Hi Peter, >> >>> >> >> >>> >> I've never had a problem with apt generating resources in the wrong >> >>> >> directory. Can you attach an example test project to demonstrate >> >>> >> the >> >>> >> problem here? >> >>> >> >> >>> >> Cheers, >> >>> >> >> >>> >> Mark >> >>> >> >> >>> >> 2009/12/17 Peter Lynch <[email protected]>: >> >>> >> > Hi, >> >>> >> > >> >>> >> > I created jira http://jira.codehaus.org/browse/MOJO-1478 and >> >>> >> > provided >> >>> >> > the >> >>> >> > patch to fix it. >> >>> >> > >> >>> >> > I thought I'd post here asking if someone is able to apply the >> >>> >> > patch >> >>> >> > for >> >>> >> > me? >> >>> >> > I tried irc first... >> >>> >> > >> >>> >> > I am offering any help in this regard. Anything from applying >> >>> >> > the >> >>> >> > patch >> >>> >> > to >> >>> >> > doing a release of apt-maven-plugin with the patch, however I do >> >>> >> > not >> >>> >> > have >> >>> >> > any commit rights for moho/codehaus. >> >>> >> > >> >>> >> > Anyone have a minute to spare? >> >>> >> > >> >>> >> > -Peter >> >>> >> > >> >>> >> > >> >>> >> >> >>> >> >> >>> >> --------------------------------------------------------------------- >> >>> >> To unsubscribe from this list, please visit: >> >>> >> >> >>> >> http://xircles.codehaus.org/manage_email >> >>> >> >> >>> >> >> >>> > >> >>> > >> >>> >> >>> --------------------------------------------------------------------- >> >>> To unsubscribe from this list, please visit: >> >>> >> >>> http://xircles.codehaus.org/manage_email >> >>> >> >>> >> >> >> >> >> > >> >> --------------------------------------------------------------------- >> To unsubscribe from this list, please visit: >> >> http://xircles.codehaus.org/manage_email >> >> > > --------------------------------------------------------------------- To unsubscribe from this list, please visit: http://xircles.codehaus.org/manage_email
