bug: breakpoints in in an anonymous class method executed on a virtual thread callstack are skipped

[Ben Berman]

Thanks for replying to this. I am not sophisticated enough to answer these
questions. Since this bug is real, the best thing would be if someone who
is more sophisticated tries the reproduction that is simpler and shared by
Vlad here:
https://youtrack.jetbrains.com/issue/IDEA-324002/IntelliJ-debugger-doesnt-break-nor-does-Eclipse-in-a-virtual-threads-project-inside-anonymous-class-methods-reproduction#focus=Comments-27-7821461.0-0
. It reproduces in:

openjdk version "21" 2023-09-19
OpenJDK Runtime Environment (build 21+35-2513)
OpenJDK 64-Bit Server VM (build 21+35-2513, mixed mode, sharing)

on Windows.

On Mon, Jul 10, 2023 at 6:52 PM Chris Plummer <chris.plummer at oracle.com>
wrote:

> One other thing to add. When you see a "JDWP exit error" like you
> mentioned, that should result in the debuggee exiting. I'm not sure how
> you are seeing the println() also being executed, unless the "JDWP exit
> error" is happening afterwards, such as when the debuggee exits on its
> own, and the debug agent is shutting down.
>
> Chris
>
> On 7/10/23 3:27 PM, Chris Plummer wrote:
> > Hi Ben,
> >
> > The first time I read this I didn't notice the link you provided to
> > the jetbrains issue you filed. I see about halfway down the page you
> > show the following exception:
> >
> > JDWP exit error AGENT_ERROR_INVALID_THREAD(203): missing entry in
> > running thread table
> > [src/jdk.jdwp.agent/share/native/libjdwp/threadControl.c:1109]
> >
> > As is pointed out later in this issue page, this looks like
> > JDK-8305209 <https://bugs.openjdk.org/browse/JDK-8305209>, which was
> > fixed a few months ago. You were ask to try a recent version of JDK
> > 21, and you replied back that it still reproduced, but the error you
> > show later indicates that it is still happening at line 1109. It
> > should be line 1116 if you are using a fixed version. Just wanted to
> > double check that line 1116 is the line number you are seeing the
> > error at when you use JDK 21.
> >
> > Also, when this error happens, it should cause the debuggee to exit
> > with a core file and hs_err file. Are either of these present? The
> > hs_err file would be very useful since it will contain a stack trace
> > at the point of the error.
> >
> > thanks,
> >
> > Chris
> >
> > On 6/29/23 2:23 PM, Ben Berman wrote:
> >> Investigating an issue in both IntelliJ and Eclipse debuggers led to
> >> this bug: I can't break inside an anonymous class method of the form
> >>
> >> ```
> >> new Object() {
> >>   void test() {
> >>     System.out.println("breakpoint here does not work");
> >>   }
> >> }.test();
> >> ```
> >>
> >> due to a buggy interaction between JDWP and virtual threads. The
> >> console statement in the example above does indeed print, i.e. it is
> >> executed.
> >>
> >> What would the next steps be to help with investigating and resolving
> >> this issue?
> >>
> >>
> https://youtrack.jetbrains.com/issue/IDEA-324002/IntelliJ-debugger-doesnt-break-nor-does-Eclipse-in-a-virtual-threads-project-inside-anonymous-class-methods-reproduction?clearDraft=true&description=%0A%0AIU-231.9161.38,%20JRE%2017.0.7%2B10-b829.16x64%20JetBrains%20s.r.o
> >> <
> https://youtrack.jetbrains.com/issue/IDEA-324002/IntelliJ-debugger-doesnt-break-nor-does-Eclipse-in-a-virtual-threads-project-inside-anonymous-class-methods-reproduction?clearDraft=true&description=%0A%0AIU-231.9161.38,%20JRE%2017.0.7%2B10-b829.16x64%20JetBrains%20s.r.o>.,%20OS%20Mac%20OS%20X(aarch64)%20v13.4,%20screens%203456.0x2234.0;%20Retina
>
> >>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.openjdk.org/pipermail/loom-dev/attachments/20230918/bfeede42/attachment-0001.htm>