Hilko Bengen wrote: > Yes, "-a" is probably broken. I still consider that an improvement over > #694176 which I intended fix in the first place -- and which blocks the > package from reaching testing. > > Having poked at the adb sources some more, I come to the conclusion that > upstream's intention is to provide two modes for setting up an adb > daemon: > > (1) A "local" TCP socket (default wrong behavior, see #694176) > (2) A TCP socket on all interfaces (even less secure, but whatever...)
Correct. adb is a debugging tool, almost never run on multi-user systems. I agree that the default behavior is insecure on multi-user systems, and switching to a per-usr socket would make sense. For example, a UNIX domain socket stored in the user's home directory, with permissions set to only allow that user to access it. Nonetheless, I don't believe it makes sense to make this change without coordinating with upstream, and I don't believe #694176 should be release-critical. The security aspect of the bug simply doesn't come up on normal single-user developer systems. So, I'd propose that the correct fix is to revert the NMU completely, upload -2 as -2.2 or -3, reopen #694176 and set its severity to "important", and then work with upstream to come up with an alternative default. - Josh Triplett -- To UNSUBSCRIBE, email to debian-bugs-rc-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
