The default handler fired for every update the command matchers didn't
claim, so plain chatter, service messages, media, and replies to the bot
all drew a "Nieznana komenda" response — spammy in group chats. Gate the
fallback on a leading bot_command entity so only real (unrecognized)
commands get the help reply.
Add tests for the previously-untested handlers (ktoSprzata, sync, unknown,
matchCommand, BotSendFunc, PublishCommands) using a real *bot.Bot pointed at
an httptest server that captures sendMessage replies. Covers the /sync
success, throttled, and failure paths plus the non-message fallback.
In group chats Telegram appends the bot addressee to the bot_command
entity (/sync@hs_dyzur_bot), and the library's MatchTypeCommand compares
the whole token verbatim, so commands never matched. Replace it with
custom MatchFuncs that read the bot_command entity span and strip any
@suffix before comparing the command name.
Resolve the current-year sheet tab per fetch instead of freezing it at
startup, so a long-running process follows the New Year rollover without a
restart. Guard against the regression this introduces: a successful fetch
that returns no duty weeks (e.g. a freshly created, empty next-year tab) now
keeps the last-good cache instead of clobbering it with empty data.
Surface a silently-stale cache — which is not a crash, so process-level
supervision can't catch it — by escalating to a loud error log once no sync
has succeeded for 6h (new Store.LastSynced accessor backs the check).
Throttle manual /sync to once per minute across the chat to avoid spamming
the Google Sheets API; the check-and-set is mutex-guarded since Telegram
dispatches each update in its own goroutine.
Update README accordingly.
These were never committed (hidden by status.showUntrackedFiles=no): without
go.mod/go.sum a fresh clone would not build, and the source files committed in
abb38e3 were missing their accompanying tests.
Kamil 'Kamaś' Bruchal