Anders Hammar wrote:
On Mon, Jan 17, 2011 at 08:48, Lukas Theussl<ltheu...@apache.org> wrote:
Ok, I get it. The xdoc should be written with the same encoding that is
used when it is read to generate the final report, which is
project.build.sourceEncoding (by default).
No, it shouldn't. The xdoc xml file should be written with the encoding
specified in the xml prolog.
Well, the xdoc is generated, so there is no xml prolog yet. But the xml
encoding declaration should of course match the actual file encoding.
I tried the latest snapshot and the prolog defines UTF-8 but the text is
written with the encoding specified by project.build.sourceEncoding. If
project.build.sourceEncoding is something other than UTF-8, non-ascii chars
are garbled if properly reading the XML (i.e. honoring the encoding
specified in the xml prolog).
Then my guess is that PluginXdocGenerator writes the wrong prolog. Can
you construct a test case? I hate encoding issues, but since I was
stupid enough to try to fix it... ;)
-Lukas
/Anders
I fixed it, thanks!
-Lukas
Benjamin Bentmann wrote:
Lukas Theussl wrote:
I am confused indeed. I simply used MPLUGIN-180 to verify it works, but
just realized it's not good because both input and output encoding are
the same there.
Right, it gets only interesting if the two encodings have different
values, e.g. setting project.reporting.outputEncoding=ISO-8859-1 in
Anders' example project will yield garbage from "mvn site".
However, the problem is not that the generated report has the wrong
encoding, rather the generated, intermediate xdoc, from which the report
is generated (run 'mvn site:site' and 'mvn plugin:xdoc' on the attached
test project and compare the generated xdoc). For writing this, I would
have thought one should use the outputEncoding parameter? Tell me if I'm
wrong and I'll change it.
The intermediate xdoc is produced by the PluginXdocGenerator from
previously extracted MojoDescriptors. It is hard-coded to use UTF-8 for
the encoding of the generated xdoc. All fine here.
"site:site" invokes PluginReport whereas "plugin:xdoc" invokes
XdocGeneratorMojo. The latter extending AbstractGeneratorMojo will use
the proper/configured encoding when reading the Java sources to extract
the mojo descriptors and eventually generates proper xdoc. But
PluginReport in version 2.6 uses the wrong encoding (platform default)
and as such produces bad MojoDescriptor instances upon source parsing.
Bad MojoDescriptor -> bad xdoc.
So the encoding issue doesn't happen on the output/report side, it
happens on the input/source side, the bad output is just a followup
error. The proper fix is to align PluginReport with the encoding
handling done in AbstractGeneratorMojo such that the source files
consistently get parsed with the same/configured encoding when calling
the mojo scanner.
Benjamin
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@maven.apache.org
For additional commands, e-mail: dev-h...@maven.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@maven.apache.org
For additional commands, e-mail: dev-h...@maven.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@maven.apache.org
For additional commands, e-mail: dev-h...@maven.apache.org
