RFR 8193216: Filer should warn if processors redefine symbols from the classpath or sourcepath

[Liam Miller-Cushon]

On Thu, Dec 14, 2017 at 8:20 PM, joe darcy <joe.darcy at oracle.com> wrote:

> Given the change in long-standing warning behavior, to be thorough I'd
> like to see a quick CSR created for this issue. (From the issue in JBS,
> under the "More" box at the top of the page selecte "Create CSR" toward the
> bottom. Fill out the form presented in the the new issue and when you're
> ready to have it looked at hit "Finalize.")
>
Thanks, I started a CSR: https://bugs.openjdk.java.net/browse/JDK-8193576

Does this require a specification change? The documentation for
-Xlint:processing isn't very specific, and the documentation for Filer
doesn't discuss warnings.

> For the test, I don't know if there would be additional coverage if a
> second compile line was added to test a pre-modules world...
>
I double-checked that the test still passes with the additional compile
line.

> I'll adjust my eventual fix for
>
>     JDK-8193462: Fix Filer handling of package-info initial elements
>
> to treat package-info files analogously with the changes from this issue
> after the changes go back.
>
Thanks!

> Perhaps separate from this change, it may be reasonable to check for
> obviously overlaps in the source locations and target locations.
>
Do you have an example of that kind of overlap? I'm not certain what source
and target mean in this context.

I've seen issues related to overlaps between the sources of a compilation
and its classpath, or between jars on the classpath. That kind of overlap
is relatively common, but an optional lint warning might be nice to have.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.openjdk.java.net/pipermail/compiler-dev/attachments/20171214/c37e13c4/attachment-0001.html>