v0.2.2

bump version

v0.2.1

bump version

v0.2.0

bump version

v0.1.9

bump version

v0.1.8

Feature/improve sdb reliability (#20)

* init reliability

* Start the service with the television, and stop polling for it

config.xml declares the service on-boot and auto-restart, so the set can be
switched on and reached from a phone with nobody in front of it. Confirmed on a
partner-signed QE65S93DATXXN: after a restart the service was up, one process,
with no loopback client ever in its log.

Two things follow from that.

The pairing code is now minted once and kept in homebrewConfig.json. One
regenerated every start would be readable only off the screen that booting the
service exists to avoid, and every reboot would unpair every phone. A developer
build still pins it to 000000 and stores nothing.

Back no longer stops the service. It is a separate application that outlives
the page now; the theme audio is still torn down so it cannot play on over the
Tizen home screen.

The television's page also stops polling. It ran three XHR loops — /pin every
500ms until the service answered, /logs every second, /state every five — and
now holds one socket instead. The recorder gained subscribers, the protocol
gained `watch` and `log`, and the greeting hands a loopback caller the pairing
code the way GET /pin always has. Startup detection is the connect attempt
itself: there is no platform event for a port opening, but a refused connection
is an event, so the retry fires on failure and backs off rather than running on
a clock. Restart drops its 200ms wait-for-exit poll for the same reason — the
socket dropping is the signal.

* Mint a Partner certificate by default

The service declares on-boot and auto-restart, and both are partner signed
only, so the pair a new set gets has to reach that far. --public keeps the
old level for accounts that cannot mint Partner.

--output now takes a copy of the author certificate it kept, so a second
pair stands on its own and a public one can sit beside the partner one
instead of overwriting it.

* Tear the sdb connection down without resetting the next client

_abandon now returns early when no stream is open, so a socket closing
after a clean shutdown stops inventing an error for nobody, and it builds
the ESDBCLOSED itself rather than having each caller carry one.

close() sends CLSE through _closeStream, which already knows a stream with
no remote id yet, and a connection still being opened is closed as soon as
it arrives instead of outliving the shutdown that asked for it.

* Re-sign native packages too

Signing hashes everything but the signatures, so a tizen-manifest.xml
package needs nothing else; the manifest was only ever read as proof the
archive is a Tizen package at all.

* Version 0.1.8

v0.1.7

bump version

v0.1.6

bump version

v0.1.5

bump version

v0.1.4

bump version

v0.1.3

bump version