Attach to Virtual Machine not working

[kwilliams]

That gave me a different result. It looks like I was able to attach. The debug windows came up and I was able to pause and see a call stack. but all of my attempts to set a breakpoint resulted in "Unable to set breakpoint at this location: Could not find the specified function".

[patrick]

Trying to narrow it down.  This may sound ridiculous, but if you detach, set the breakpoint at that location, and attach, does it stop at the breakpoint at all?

[patrick]

I've been trying the debugger with some different codebases than what I usually test with, and I see that breakpoint message with a location in SBT.  My changes for checking the stepping speed and locals had a few more rough edges that I wanted, so there are also some issues I see related to that and not inherent in the real fix.  Like double clicking on a call stack location to go to that location not always working.  (even though a breakpoint at that location would locate the file).     Feel free to uninstall the test install, and put back the regular one.

This isn't hot-fixable, which is why I've been having to send jars and installers.  I'm working on getting these in for 26.0.1, which will come out before the end of the year. 

[kwilliams]

Ok, thanks.

I tried playing around with it some more. It's been timing out and failing to connect.

I'll wait to hear from you.
We're pretty eager to move to SlickEdit, so let me know if you have something to test or need more info.

[patrick]

I think the last debug logs with timeouts we got, it was the one where it was started with some of the old code still running with your regular config, so I don't think that's the same problem.

If you can reproduce it without too much trouble, you can clear the debug.log, set the def-debug-logging var, and gather the debug.log and send it.  And the output in the Build window, though there shouldn't be much there with the updated installer. 

Without the logs I'm assuming it's in the borderline timeout territory because of the problem where it gathers too much information on connect.  That's definitely on the list to fix for the point release.

[kwilliams]

Hi Patrick,

any word on releasing a fix for this?

[patrick]

Not yet.  I had planned for it to be in 26.0.1, but we ran into other issues that ate into my time.  So I'm tentatively aiming for 26.0.2.

[patrick]

The fixes for this did go in to 26.0.2.   Some notes:

1) The attach should be faster, I reorganized some of the initialization code whose time was linear with the number of classes the JVM has loaded.  There still is a linear dependency, this just greatly reduces the constant factor associated with the setup.
2) The problems with breakpoints and file prompts related to the directory structure not matching the package should be gone now.
3) Should no longer see mangled names for some closure parameters.