Hi Roman,
TL;DR: The behavior is as expected.
If nothing else, note that `--source` describes the language level of
the input source code, and has no relationship whatsoever to HTML
version used for the output. This is somewhat similar to javac, where
`--source` specifies the version of the input source files, and
`--target` specifies the classfile version of the generated class files.
There are other major differences between JDK 8 javadoc and more recent
versions. More obvious than the version of HTML, there is the
difference in the page layout: JDK 8 javadoc uses frames and predates
the Search facility; later versions do not use frames, and provide the
Search mechanism and more in-page links to aid with navigation.
The issue of the form of anchors generated by different versions is an
interesting special case, with no easy solution at this time, other than
to use compatible versions of javadoc for the API being cross-linked.
There is no information available to the `-link` or `-linkoffline`
options to know the form of anchors used in the documentation being
linked to.
That being said, there is ongoing work to standardize _and specify_ the
form of anchors and major structural elements to mitigate against
problems going forward.
-- Jon
On 11/18/22 7:32 AM, Roman Marchenko wrote:
Hi,
As far as I know, Html4 support was removed from Javadoc a time ago.
So now Javadoc generates html5-like anchors even with “-source 8” or
“-release 8” option specified. There is JDK-8187521 added
TestAnchorNames.testHtml5 which checks if Html5-like anchors are
generated for “-source 8” option.
On the other hand, Java8’s Javadoc still supports html4 and doesn’t
support html5. So it turns that docs generated by Java8’s Javadoc are
not compatible with docs generated by any other (I guess starting from
jdk17 and above) “Javadoc -release 8” because they use different types
of anchors. So it’s impossible to use cross links in such cases.
I haven’t found any discussion explaining that, so could anyone
explain, please, if it is correct behavior or not?
- Roman
