gdb-7.8 branching status (2014-06-04)

Joel Brobecker brobecker@adacore.com
Thu Jun 5 21:03:00 GMT 2014


> >   . I also noticed what looks like some failures with gdbserver
> >     on Windows, but I haven't had time to look into this.
> >     I am hoping it's the same issue was on LynxOS.
> 
> I can now say with relative certainty that it isn't the same issue.
> It looks related to the fact that target-async is enabled by default.
> I'm still gathering evidence, but the the reproducer is a little racy;
> it doesn't always fail, and doesn't always fail at the same point :-(.
> 
> Sadly, I also noticed that the "maint set target-async" command is
> missing on Windows hosts :-(. So that made testing my theory a little
> more difficult, and added yet more work before we can cut the 7.8
> branch. But in any case, manually setting target_async_permitted
> and target_async_permitted_1 (I am paranoid) to 0 allowed me to
> pass my test 20 times in a row, as opposed to having it fail roughly
> 12-15 times.

Hmmmpf, really elusive one to track down as the problem becomes more
and more elusive as you add traces. It looks like communication issue,
although I don't have all the details yet. I created a PR:

    GDB+GDBserver hangs on Windows waiting for stop event since target-async
    on by default
    https://sourceware.org/bugzilla/show_bug.cgi?id=17028

I mentioned target-async in the subject since this seems to be
the trigger that allowed the problem to show up, but so far,
I don't see why this would make a difference.

-- 
Joel



More information about the Gdb-patches mailing list