feat: support new device ID format #1911
Conversation
andyholmes
force-pushed
the
andyholmes/device-id-verification
branch
4 times, most recently
from
6fae25a to
3613d85
Compare
| _migrateConfiguration() { | ||
| if (Device.validateId(this.settings.get_string('id'))) | ||
| return; | ||
|
|
||
| // Remove the old certificate, serving as the single source of truth | ||
| // for the device ID | ||
| try { | ||
| Gio.File.new_build_filenamev([Config.CONFIGDIR, 'certificate.pem']) | ||
| .delete(null); | ||
| Gio.File.new_build_filenamev([Config.CONFIGDIR, 'private.pem']) | ||
| .delete(null); | ||
| } catch { | ||
| // Silence errors | ||
| } |
There was a problem hiding this comment.
This all looks sensible, but I wonder if we need some sort of notification to the user, when the migration process triggers? ("Hey, we had to re-generate your local device ID, so all of your devices need to be re-paired. Lusciously sorry!")
|
I hate this test suite so much, so much. |
|
I don't mind admitting, it scares the bejesus out of me. I managed to wrap my brain around JUnit1 in a day or two, a while back, at least well enough so that I could accomplish what I needed to. But jasmine? Uh-uh. (And the friction increased by a factor of about 5, once the tests could no longer be run un-installed after the move to ESM.)
|
andyholmes
force-pushed
the
andyholmes/device-id-verification
branch
3 times, most recently
from
9384e5b to
de04c0f
Compare
|
|
andyholmes
force-pushed
the
andyholmes/device-id-verification
branch
3 times, most recently
from
9bd75d9 to
48d0e80
Compare
Add support for the planned changes in protocol version 8. See: https://invent.kde.org/network/kdeconnect-android/-/commit/454b2fd5d7b3162443fbec4a6f0b7cafa1b554ee
Update the certificate generation to produce EC-based certificates with the required device ID format.
Check incoming device IDs and reject connections from those with
an ID not matching `/[a-zA-Z0-9_]{32,38}/`.
Migrate installations to new certificates, with device IDs following the protocol v8 constraints.
Add a brief notification, explaining that GSConnect has been updated and some devices may require re-pairing.
andyholmes
force-pushed
the
andyholmes/device-id-verification
branch
from
48d0e80 to
59be25a
Compare
|
There's still some bits left to do for protocol v8, so I'm probably going to split this up before review. |
Update the certificate generation to produce EC-based certificates
with the required device ID format and reject connections from those
with an ID not matching
/[a-zA-Z0-9_]{32,38}/.Migrate installations to new certificates, with device IDs
following the protocol v8 constraints.
closes #1910