Many hyperlinks are disabled.
Use anonymous login
to enable hyperlinks.
50 most recent check-ins
|
2025-01-29
| ||
| 11:46 | (build) more c23 build warnings Leaf check-in: 28f7bcf113 user: void tags: master, trunk | |
| 09:43 | (build) fix warnings from testing -std=c23 check-in: 81ecabbba0 user: void tags: master, trunk | |
| 09:34 | (build) bump version check-in: 87fe0bf5e8 user: void tags: 0.7.0.1, master, trunk | |
| 09:34 | (net) start on passive-synch Passive discovery just prints out the known tags discovered but if the IP is different there is no way to synch to the keystore as of yet. This is problematic as there can be many hosts to one tag, and each has (potentially) a different keypair. This means that in order to update the keystore we would have to probe every item in the tag until we find one that authenticates, and then update the IP, if one is even used. If it's host-based and the external naming system is out of date (DNS TTL ..) changing it may have adverse effects so it should only apply to hardcoded IPs (which LAN discovery was meant for anyway since the tag is a naming system). check-in: 82017f571b user: void tags: master, trunk | |
|
2025-01-28
| ||
| 14:25 | (net) add .monitor permission for debugging controller check-in: 39b7722b5e user: void tags: master, trunk | |
| 14:12 | (tui) lua-bindings, switch handling of nested callbacks There are too many possible chain-invocations of handlers from widget that just forwarding the right user-data based on calling context is too fragile. Switch to extracting the tui- userdata for any window handler regardless of origin. This should fix Luajit crashes (which is more sensitive to unbalanced stack on function exit). check-in: 85b1da5a6f user: void tags: master, trunk | |
| 13:27 | (lua) dump_stack debug format nit check-in: d2cf25f7eb user: void tags: master, trunk | |
|
2025-01-27
| ||
| 17:36 | (lua) read_rawresource() stack index oob trigger for LuaJIT check-in: df8f4b1b59 user: void tags: master, trunk | |
| 08:35 | (doc) minor nits check-in: 38f7d92b39 user: void tags: master, trunk | |
| 00:48 | (lua) _fatal handler didn't get passed 'msg' after monitor refactor check-in: 89c13b0e56 user: void tags: master, trunk | |
|
2025-01-25
| ||
| 13:21 | (lua) interactive error reporting regression dropped msg check-in: d843224e2b user: void tags: master, trunk | |
|
2025-01-24
| ||
| 19:55 | (build) last merge didn;t pull in header check-in: cdcae4ae23 user: void tags: master, trunk | |
| 11:01 | (build) fix fork from old OBSD device commit check-in: aa14090e07 user: void tags: master, trunk | |
| 09:09 | (build) OpenBSD+Cmake GLVND bug workaround check-in: a80ccf47a1 user: void tags: master, trunk | |
|
2025-01-22
| ||
| 13:44 | (build) openBSD build fixes check-in: e94737baf7 user: letoram tags: master, trunk | |
| 10:41 | (net) add discovery_beacon to dirsrv config This will spawn a thread beaconing when the directory server and the option has been enabled in the config. It's not the default. check-in: 7ba6b6efb1 user: void tags: master, trunk | |
| 08:13 | (net) permit beacon and passive discovery on the same host This goes through the painful dance of checking if the source addr of the beacon match the interface we receive on, and if so, skip triggering on_unknown or on_beacon. check-in: 28b3fcf45c user: void tags: master, trunk | |
|
2025-01-19
| ||
| 10:43 | (net) safeguard against served client polluting epipe check-in: 5800861a8e user: void tags: master, trunk | |
| 10:43 | (lua) have attached monitor switch out the sigusr1 handler on attach/detach check-in: dfc094b12f user: void tags: master, trunk | |
| 00:03 | (core) add 'out' and 'detach' commands to monitor, 'monitor' key to arcan ns arcan_db add_appl_kv arcan monitor "/path/to/somescript.sh" somescript.sh: export ARCAN_CONNPATH="a12s://dd@" export ARCAN_STATEPATH="/path/to/keystore" export ARCAN_ARG="cli=lua:args=builtin dev:args debug attach arcan $1" export HOME=/home/myhome afsrv_terminal unlink "$1" this would cause a script error in the Lua VM to spawn Lash#Cat9 with the debug dev builtin, attaching to a FIFO created by the monitor. The builtin would, in turn, send the 'out' attachment and connect that to another FIFO (avoiding the ordering problem of establishing both in and out through the same process). When the debugger is done, it sends the 'detach' command and the FIFOs are unmapped. check-in: 7c7ba007c3 user: void tags: master, trunk | |
|
2025-01-18
| ||
| 23:59 | (platform) add platform_monitor_launch (partial) This prepares for letting the monitor launch and attach an external debugger on-demand. It's incomplete insofar as we need a real mksfifo version rather than a hardcoded name, but the feature is still being tested. check-in: 9314bc55a3 user: void tags: master, trunk | |
| 23:57 | (tui-lua) add fifo-in as :fopen() mode check-in: 3688c57832 user: void tags: master, trunk | |
| 23:55 | (lua) wrong format for traced key in output check-in: b5a19744e2 user: void tags: master, trunk | |
| 23:55 | (lua) compiler warning on wrong int type in printf check-in: bd80e65182 user: void tags: master, trunk | |
| 23:54 | (net) committed the wrong patch to a12.c (see prior commit) check-in: 6c548f3705 user: void tags: master, trunk | |
| 23:42 | (lua) open_nonblock should give precedence to UD on mismatch This is an API design oversight that lead to the confusion around BCHUNK_IN/OUT. Take the case: open_nonblock(netvid, false, "somefile") -> ud_in decvid = launch_decode() open_nonblock(decvid, true, "*", ud_in) For routing a file from net to decode. On the net-side we receive a BCHUNK_IN that gets converted to a BCHUNKSTATE request and the server would enqueue the bstream on the channel. We receive that and convert to a BCHUNK_IN that gets bound to ud_in in read mode. Now the last call is conflicting, it would result in ud_in descriptor transferring to decvid but decvid would get BCHUNK_OUT. Since we can't break the prototype for this function, ignore the direction request and use whatever matches ud_in. check-in: 740bba9438 user: void tags: master, trunk | |
|
2025-01-17
| ||
| 00:09 | (net) two regressions caused by outbound- prefix This should let afsrv_net connect by @tag go back to working (outbound prefix wasn't set and then not present in accepted keys) as well as ARCAN_CONNPATH a12s://@tag to register source again (same problem). check-in: ca59b2d2a9 user: void tags: master, trunk | |
|
2025-01-15
| ||
| 22:43 | (shmif) change how RESET delay slot is activated This should make crash recovery more robust for clients (looking at you Xarcan) where: arcan_shmif_signal(sigvid) <- server crashes at this point -> _poll() latched to epoll triggering (but descriptor is dead) -> epoll never firing dispatch event -> _RESET never gets consumed check-in: de67c4fa06 user: void tags: master, trunk | |
| 06:36 | (lua/a12) flip to IN/OUT should've been in a12.c not nbio.c check-in: 0be31760fe user: void tags: master, trunk | |
|
2025-01-14
| ||
| 23:25 | (shmif) move structure definition around to silence clang++ check-in: f155fc5301 user: void tags: master, trunk | |
| 22:56 | lua: implement breakpoints check-in: cd490758b3 user: void tags: master, trunk | |
|
2025-01-13
| ||
| 07:43 | (lua) monitor trace navigate through hierarchy of tables check-in: d12b363f2c user: void tags: master, trunk | |
|
2025-01-12
| ||
| 23:42 | lua: start work on monitor command for dumping table check-in: 2a84c29962 user: void tags: master, trunk | |
|
2025-01-11
| ||
| 19:22 | (lua) trace name field wrong check-in: 7427b14210 user: void tags: master, trunk | |
| 11:03 | (core) default to stdout if monitor-in is set but not out check-in: e2b2fa5e21 user: void tags: master, trunk | |
| 09:48 | (core) change how -C works to allow it to specify path to fifo This breaks tools that relied on the stdin default behaviour, that should only be the cat9 arcan debug tool and arcan-net (which gets changed with this patch). check-in: 55404941be user: void tags: master, trunk | |
| 09:11 | (arcterm) expose ARCAN_ARGS=args=hi:args=there to tui.arguments check-in: 39752ac23f user: void tags: master, trunk | |
| 09:11 | (core) add 'paths' command to monitor to show namespace mapping check-in: 0e8926fe80 user: void tags: master, trunk | |
|
2025-01-10
| ||
| 19:20 | (lua) trigger monitor dump on hooked call, forward entrypoint name check-in: e283d7c68a user: void tags: master, trunk | |
|
2025-01-09
| ||
| 20:03 | (lua) fix regression in _fatal handler and pcall check-in: dc77de97e3 user: void tags: master, trunk | |
|
2025-01-08
| ||
| 23:21 | (lua) swap out error-handler for monitor-if dumper in -M mode check-in: 94c2511757 user: void tags: master, trunk | |
|
2025-01-07
| ||
| 03:01 | (lua) add entrypoint hooking controls for monitor This lets the monitor set a mask of entry-points to break on when reached. check-in: a46f5d36c9 user: void tags: master, trunk | |
|
2025-01-06
| ||
| 04:14 | (lua) allow warning to trigger monitor watchdog (attached debugger) check-in: 36d0b5baaf user: void tags: master, trunk | |
|
2025-01-05
| ||
| 16:28 | (lua) forward backtrace data check-in: 5796f9f227 user: void tags: master, trunk | |
|
2025-01-04
| ||
| 16:10 | (lua) change alt_call to convey a trigger name This is intended for automatic breakpoints on entrypoints into the Lua VM. This defines a large bitmap of possible tracesources that we then map to upcoming breakpoint command in the monitor interface. That would then set the mask of active triggers, and when the mask matches, change the lua trace function to break on the first instruction in the entrypoint. check-in: 5aa80a095c user: void tags: master, trunk | |
| 03:12 | (lua/trace) make output match shmif-pack check-in: fb4e83e6d8 user: void tags: master, trunk | |
|
2025-01-03
| ||
| 17:42 | (lua/monitor) change print if no tracer is attach, implement cmd_eval check-in: 2b34283343 user: void tags: master, trunk | |
|
2024-12-31
| ||
| 11:51 | (lua) start wiring up internal debugger check-in: 4a8ba7a150 user: void tags: master, trunk | |
| 04:15 | (doc) current state of a12 covered check-in: 733170096b user: void tags: master, trunk | |
|
2024-12-30
| ||
| 01:20 | (doc) a12.md, nearly covered everything check-in: 3834e948ad user: void tags: master, trunk | |