The buggy behavior was that the 'adb backup ....' host command line
invocation would hang forever, even after the backup finished and
the forked device-side subprocess had been reaped. The reason for
this is that the device adbd end of the socketpair used to send
the data back from the forked subprocess was still stuck readable
even after the remote end of it had been closed.
With this patch, the thread whose job it is to waitpid() in order
to harvest the forked child process also closes the local (adbd)
end of the socketpair. This makes the fdevent logic notice that
the socket is dead, at which point it cleans up everything including
the communication back to the host.
Change-Id: I90e7e4e5db36c5a6f7363708b29a6d2c56d1250e
* Increase transfer buffer size to 32K
* Add logging about error conditions and fd teardown
* Pass the fd number as a command line option to the 'bu' subprocess
* Properly harvest the 'bu' subprocess after it's done
Change-Id: Id44dde25778ecf43c5604fd9d01d726ba58861e5
It won't actually do anything until the 'bu' tool and framework are
updated to respond properly, but this is the adb side of the
necessary infrastructure: we copy the tarfile into the socket pointed
at the device, using the existing mechanisms.
Change-Id: Ic3b5779ade256bd1ad989a94b0685f7b1a7d59d2
