diff options
| author | user <user@clank> | 2026-07-17 19:47:12 +0200 |
|---|---|---|
| committer | user <user@clank> | 2026-07-17 19:47:12 +0200 |
| commit | 7efcb26671563ea8f734e752e14a04455ecfd663 (patch) | |
| tree | 9f7b2af00360b81fffeb22bc72a2d8123bed2917 /src/net.asm | |
| parent | gbhub: make GB drops attributable, not spooky (diff) | |
| download | gbos-7efcb26671563ea8f734e752e14a04455ecfd663.tar.gz gbos-7efcb26671563ea8f734e752e14a04455ecfd663.tar.xz gbos-7efcb26671563ea8f734e752e14a04455ecfd663.zip | |
docs: link-drop investigation resolved - drops are self-healing now
Rewrite the notes from 'open / not root-caused' to the resolution: the
socket never mysteriously dies; every producible drop signature is the
emulator process dying silently (SIGPIPE on console echo after a hub
restart, terminal death/HUP/stray quit keys in windowed mode - which
exactly explains the 'only when idle' correlation - or a guest $ED
poweroff), plus observation artifacts in the original session's
tooling. Records the fixes shipped on both sides (emulator link
auto-reconnect + SIGPIPE ignore + $ED logging; hub eof-vs-error
logging, kept stderr, EBUSY guard), the honest spawn-mode caveat, and
what's still worth doing.
Diffstat (limited to 'src/net.asm')
0 files changed, 0 insertions, 0 deletions