aboutsummaryrefslogtreecommitdiffstats
path: root/src
diff options
context:
space:
mode:
authoruser <user@clank>2026-07-17 23:20:55 +0200
committeruser <user@clank>2026-07-17 23:20:55 +0200
commit15b4767b17f78a827fd221e941c7ac7aa987e02f (patch)
treed9f7a414a1b9dd3d88bc940c60538c2f977423ff /src
parentirc: strip mIRC formatting, drop 004/005 noise, fix a RAM-bank-swap landmine (diff)
downloadgbos-15b4767b17f78a827fd221e941c7ac7aa987e02f.tar.gz
gbos-15b4767b17f78a827fd221e941c7ac7aa987e02f.tar.xz
gbos-15b4767b17f78a827fd221e941c7ac7aa987e02f.zip
irc: keep the keyboard live during RX floods (poll input every pass)
'/join #sl0p right after connect' looked broken: the main loop only polled input when net_recv_nb returned nothing, and 'continue'd on every received segment. During a big MOTD (irc.sl0p.foo's is ~60 lines of color art) that starved input for the whole ~20s flood - the OSK wouldn't even open, and keystrokes were dropped. gbtype input fared a bit better (buffered in the 64-byte console ring) but still wasn't serviced until the flood ended. Poll pollin()/pollcon() once every loop iteration regardless of RX, and only idle-sleep when a pass both received nothing and read no key. The protocol/join logic was already correct (verified: JOIN #sl0p is sent, accepted, and the client switches to [#sl0p] with topic + names) - this just makes typing responsive while messages are streaming. Verified against the live server: injecting /join #sl0p *during* the MOTD flood now joins immediately instead of waiting it out. (Note: '#' is on the OSK - third row, second from the end: 10,8,":;,=+*_!?()[]<>@#~" - just not obvious.)
Diffstat (limited to '')
0 files changed, 0 insertions, 0 deletions