I am using dbus 1.2.14 with glib bindings on embedded linux (using uclibc) with mips cpu.
Occasionally (1 in 200) there is a delay of 25secs. This is due to the poll for a response timing out. In fact the response has actually arrived before the poll was issued.
In _dbus_connection_block_pending_call, it seems that it is possible for the response to arrive after the call to check_for_reply_and_update_dispatch_unlocked
but before the call to poll within _dbus_connection_do_iteration_unlocked.
It seems that no mutex is taken at this point so such a thing is possible
Upgrade to 1.2.24 and try again.
Maybe this is the same as in: https://bugs.freedesktop.org/show_bug.cgi?id=27811 ? you could try that patch.
Is this reproducible with a current stable release? (At the time of writing these are 1.2.26 or 1.4.6)
I am facing a similar issue while using dbus 1.2.16. I have also tried applying the patch specified in https://bugs.freedesktop.org/show_bug.cgi?id=27811 with no success.
Can you please let me know if this issue is expected to be resolved by moving to 1.2.26?
(In reply to comment #4)
> Can you please let me know if this issue is expected to be resolved by moving
> to 1.2.26?
We are hoping that, yes.
But if you've applied the patch and it still doesn't work, it might be a different, still-uncorrected issue.
Anyway, we're not supporting D-Bus 1.2 anymore, except for security fixes. Please try the latest 1.4 or 1.6.
> (In reply to comment #4)
> We are hoping that, yes.
Update: I can confirm that the issue is still seen after moving to dbus 1.2.26.
(In reply to comment #6)
> > (In reply to comment #4)
> > We are hoping that, yes.
> Update: I can confirm that the issue is still seen after moving to dbus
Ok, thank you. But unfortunately we don't support that release anymore. The bug might be real, but it won't get fixed there.
If it still exists in 1.6, we'll fix. In 1.4, if it's a threading issue, we will probably not fix.
Thanks. I will move to 1.6.4 and check if it resolves the issue.
Resolving WORKSFORME, please reopen if you can reproduce this on a modern dbus (and preferably provide a test case or a patch).