Change in behavior of javax.tools.JavaCompiler in JDK 8 running with -proc:only ?

[Jonathan Gibbons]

This sounds like a possible bug, but I doubt it is related to -proc:only.

In JDK 8 javac, diagnostics are classified as "recoverable" or not, 
depending on whether a diagnostic could conceivably go away if an 
annotation processor is run. It is possible that "package does not 
exist" is not being marked as a recoverable error. The other possibility 
is that the message is coming out of the last round of anno processing 
(before -proc:only) is kicking in.

-- Jon




On 03/12/2014 03:00 AM, Boaz Nahum wrote:
> Thank you so much for your reply.
>
> The simplest file that fails to compile is like this:
>
> package wf.auto_testing_jsystem_tiny_test.tint_test;
>
> import amat.jsystem.bbs_params.annotation.BuildingBlock;
> import amat.jsystem.bbs_utils.infrastructure.BuildingBlockContext;
>
> public class TinyTest {
>
>     /**
>      * A dummy comment
>      */
>     @BuildingBlock
>     public static void loadBatchFromFile(BuildingBlockContext context, 
> String batchPath) throws Exception {
>
>     }
> }
>
>
> one of the error messages  is:
>
> *     ....\TinyTest.java:4: error: package 
> amat.jsystem.bbs_utils.infrastructure does not exist
> *
> *     import amat.jsystem.bbs_utils.infrastructure.BuildingBlockContext;
> * ^
>
> As I mentioned in my question, it is expected error, because we don't 
> supply all the necessary compilation dependencies(we can't) .
>
> My question -again-, is That in JDK 7, it works perfect:
> The errors above are reported only via 
> javax.tools.DiagnosticListener#report (not printed to the console)
> javax.tools.JavaCompiler.CompilationTask#*call returns true*.
>
> In JDK 8,
> Same errors are reported both via 
> javax.tools.DiagnosticListener#report and are printed to the console,
> javax.tools.JavaCompiler.CompilationTask#*call returns false*.
>
> *In JDK 7*, If I omit '-proc:only', then I get the exact behavior *as 
> JDK 8*, that why I'm wondering, *maybe JDK 8 somehow ignore '-proc:only'*
>
> Thank again
> Boaz
>
>
>
>
> On Tue, Mar 11, 2014 at 8:24 PM, Jonathan Gibbons 
> <jonathan.gibbons at oracle.com <mailto:jonathan.gibbons at oracle.com>> wrote:
>
>     Boaz,
>
>     In JDK 8, javac is more aggressive about reporting errors (and
>     stopping the compilation) if the errors cannot be fixed by running
>     annotation processors. This should not be an issue if you are
>     using anno processors for generating code.
>
>     What sort of errors are you seeing in your compilation?
>
>     -- Jon
>
>
>     On 03/11/2014 08:27 AM, Boaz Nahum wrote:
>>     Hi
>>
>>     I hope some can give me hint how anlyse our probelms with JDK 8.
>>
>>     We use some code like this to do annotation processing (for
>>     generating code):
>>
>>     JavaCompiler.CompilationTask task =
>>                     javaCompiler.getTask(null, fileManager,
>>     diagnostics, optionsAsList, null, compilationUnits);
>>
>>                 BuildingBlockParamProcessor processor = new
>>     BuildingBlockParamProcessor();
>>                 List<Processor> processors = new ArrayList<Processor>();
>>                 processors.add(processor);
>>                 task.setProcessors(processors);
>>     if(!task.call()){ .... }
>>
>>     The code we try to process won't be compile in regular
>>     environment - a lot of stuff is missing from classpath.
>>     But that is OK, we use '-proc:only' and we *thought* that this
>>     what allows us to ignore the errors.
>>
>>
>>     When we moved to JDK 8, the exact code fails,
>>     javax.tools.JavaCompiler.CompilationTask#call returns false.
>>
>>     *My question, is a possibilty that '-proc:only' is ignored in JDK 8 ?
>>     *
>>     *Or maybe there is by-design change regarding annotation
>>     processing that we need to be aware of ?
>>     *
>>     I want to mention that if the early EAP of JDK 8 , this code run
>>     without any problems.
>>
>>     (Moving to annotation processing during compilation is big effort
>>     for us)
>>
>>     Any tip/hint will be appreciatively accepted
>>
>>     Thanks
>>     Boaz
>>
>>
>>
>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.openjdk.java.net/pipermail/compiler-dev/attachments/20140312/21f2d2de/attachment-0001.html>