Many hyperlinks are disabled.
Use anonymous login
to enable hyperlinks.
50 most recent check-ins
|
2026-09-09
| ||
| 05:45 | (net) last bits of uncompressed .hook transfers This also lets us consider more things for being sent uncompressed, e.g. check the initial bytes of a file for the zstd magic identifier. The next optimization to reduce the load for a directory server would be to offline- store a zstd version of each appl on scan/update with checksum calculation which should bring the startup times for heavy appls on a slow server down substantially. One thing with this patch to keep an eye on is that the a12 part for reading bytes for the EINTR path. Previously we had the condition: if (nts == 0 && (die || !node->zstd)) which >should< be wrong but the branch path is complicated. The new version follows the logic: we have nothing to send. the backing store is dead. it's uncompressed. (compressor might have data to flush even though the backing source is broken). Leaf check-in: dcdaa5c225 user: letoram tags: trunk, master | |
|
2026-09-08
| ||
| 22:10 | (net) modify hook_resource to specify if the transfer should go through uncompressed There are a lot of layers to this, this commit still doesn't enable the feature but adds all the plumbing needed (i.e. let hook_resource specify the preference; track it in the hook-table; communicate the preference to the worker; encode it in the type-field for the a12 transfer setup; respect it in the queued node allocation). check-in: d4c7840431 user: letoram tags: trunk, master | |
|
2026-09-07
| ||
| 01:51 | (build) include a12dir_patch helper script for arcan-net based submissions check-in: 46924ab323 user: letoram tags: trunk, master | |
| 01:50 | (net) edge-case for .resource hooks in global ns check-in: 581871b931 user: letoram tags: trunk, master | |
| 01:12 | (core) render_text on invalid/broken font shouldn't try and update cache entry for inode check-in: 04af8d48ff user: letoram tags: trunk, master | |
|
2026-09-01
| ||
| 15:37 | (decode) error fallback on static file processing regression check-in: 540a5c36d6 user: letoram tags: trunk, master | |
| 15:34 | (lwa) forward displayhints with tagged external reference IDs check-in: bb9a7cf05f user: letoram tags: trunk, master | |
| 15:34 | (doc) add missing entrypoints to generation hdr check-in: bbde5e7746 user: letoram tags: trunk, master | |
| 13:35 | (platform) fix incorrect mapping of envvars to namespaces Credit-to: Felix Uhl check-in: 941dfd2fd5 user: letoram tags: trunk, master | |
| 12:53 | (build) Fix Debug-Builds with -DLUA_TRACE_METHOD=stderr and coverage Credit to: Felix Uhl check-in: 754b681a4c user: letoram tags: trunk, master | |
|
2026-08-20
| ||
| 23:03 | (doc) remove outdated links to the github repo check-in: 68df3f5b03 user: vimpostor tags: trunk, master | |
| 22:49 | (doc) transfer wiki This salvages the last remaining pieces from the Github repo. While some documentation is dated, there is still lots of worthwhile reading material to be found here and it would be a shame to leave this behind. check-in: 4dd0eba808 user: vimpostor tags: trunk, master | |
|
2026-07-30
| ||
| 10:55 | (lua) forward viewport from arcantarget_hint, embed tested against durden check-in: ede3a220b3 user: letoram tags: trunk, master | |
|
2026-07-29
| ||
| 12:30 | (lua) first part of handover_alloc() bringup This only works from the arcan-in-arcan (lwa) context. Its purpose is to let an appl allocate a decode window for embedding (like in cat9) or discrete externally managed window with hints about WM behaviour. What's missing in this commit is the DISPLAYHINT routing and VIEWPORT event synthesis necessary for embedding control and feedback. check-in: 02cad4d2f4 user: letoram tags: trunk, master | |
| 12:24 | (doc) update changelog with API changes check-in: d616798a65 user: letoram tags: trunk, master | |
|
2026-07-26
| ||
| 01:32 | (lua) open_nonblock (write=true) struction .wrmode value change When hitting the create_new_file path, the write mode is still set to O_RDWR while it should be O_WRONLY. This normally doesn't matter, but when using open_nonblock again with the (fsrv_vid, ..., nbio_ud) form and fsrv_vid should take over the descriptor, it meant that the wrong BCHUNK_IN/OUT got sent. When this was used with arcan-net it meant that if the file was created, the pairing request turned into upload rather than download. check-in: 1b2ece89d4 user: letoram tags: trunk, master | |
| 01:29 | (decode) use SHMIF_RGBA macro for nanosvg This got undone last time we vendored the nsvg implementation. check-in: 09e1560130 user: letoram tags: trunk, master | |
|
2026-07-24
| ||
| 15:20 | (net) fix worker null-deref on message while waiting for bchunk ack This was a fun one. When the queue-events+request_resource was refactored from working only with the parent process to also be used with the ctrl- runner the dispatch for pending events needed to extract the shared state structure stored in the user portion of the shmif cont. Normally this is quick enough that there aren't any events coming from the user at the same time, but while preparing file attachments to messages we do send both a message and the file, as well as the controller broadcasting the message back, triggering this path. check-in: d149e26d64 user: letoram tags: trunk, master | |
|
2026-07-22
| ||
| 12:47 | (build/doc) minor nits check-in: 40a678adc6 user: letoram tags: trunk, master | |
|
2026-07-21
| ||
| 22:20 | (decode) add optional support for lottie vector animations This took quite some rework as the format requires a different event loop to deal with animation timing, so do_file now sets a fptr based on the format it is currently processing. check-in: 0cb264177a user: letoram tags: trunk, master | |
|
2026-07-13
| ||
| 20:40 | (net) transfer completion now forwarded to entrypoint check-in: fde780c14f user: letoram tags: trunk, master | |
| 20:14 | (shmif) evpack fix to seekframe check-in: 72a8605b65 user: letoram tags: trunk, master | |
| 01:11 | (a12) set the bchunkstate size field correctly check-in: 86f1d7690d user: letoram tags: trunk, master | |
| 01:10 | (net) expose more metadata in _store ep for ctrl This also adds a _store_status entrypoint for notifying about completion status and progress for a store, still work in progress. check-in: 7b581b2f39 user: letoram tags: trunk, master | |
|
2026-07-11
| ||
| 22:47 | (build) accidentally left SCRUB enabled in tip check-in: 5a0807ecfc user: letoram tags: trunk, master | |
| 22:23 | (shmif) retain the 'unset-category is EXTERNAL' assumption in pack/unpack There are a lot of shmif clients that doesn't set category by default, and regular shmif paths assume EXTERNAL in such cases, so evpack helper should as well. check-in: fed2408a60 user: letoram tags: trunk, master | |
| 18:10 | (shmif/a12) change arcan_shmif_evpack/unpack to match a12.md spec There can still be dragons here, as we finally move away from the previous IB hack. For testing purposes it is possible to (manually) define SCRUB_EVENT to force all enqueue/dequeue to go through a pack/unpack operation. This allows for testing without having to use arcan-net as an intermediary, as well as a quick reproduction when some pack/unpack mismatch is found. ARCAN_SHMIF_DEBUG now only logs signals and bufferstream transfers if it has been set to verbose (=2). check-in: 248079253f user: letoram tags: trunk, master | |
|
2026-07-10
| ||
| 01:05 | (net/lua) modify open_nonblock pairing and propagation to retain size This is the next step (still incomplete) in trying to assist ctrl development such that it can factor in size in determining if a _store request should be permitted or not. check-in: 933b998221 user: letoram tags: trunk, master | |
|
2026-07-08
| ||
| 13:44 | (shmif) de-union size/ns in ext.bchunk This is needed to map more metadata in the bchunkstate that gets passed when preparing an a12-define-binary-stream. We have enough space for this without breaking the hard sizeof(arcan_event) == 128, but not enough to also carry the checksum. The reason we have to somehow get this data across is that it gets hard to let the ctrl side of an appl to make an upload- decision based on size alone. While we could simply set a cap to the open_nonblock state table, that is not ideal as hitting the limit would happen after the fact, wasting data transfers. The checksum is less important in the sense that since we do have access to it in the bstream header and can make deduplication / early-completion decisions when the transfer is initiated. check-in: 92aff79f54 user: letoram tags: trunk, master | |
|
2026-07-03
| ||
| 14:46 | (net) modify setkeys to accept a wider range of values check-in: 9616ecbfef user: letoram tags: trunk, master | |
|
2026-07-01
| ||
| 15:48 | (shmif) actually mark event slot on queue saturation check-in: d3244de982 user: letoram tags: trunk, master | |
| 15:37 | (net) add --log-msg fn argument for dir-dev This is to help split out logspam versus application net- message logic. check-in: e37e03b0ca user: letoram tags: trunk, master | |
|
2026-06-30
| ||
| 20:41 | (build) header re-order fix for gcc14.3.1 + musl check-in: d105e08e6d user: btrach tags: trunk, master | |
| 20:40 | (terminal/tsm) wire in normal tsm_screen scroll ops for TUI_ALTERNATE Should fix htop check-in: 884e622736 user: btrach tags: trunk, master | |
| 20:39 | (terminal/tsm) complete CSI SGR fix check-in: 67171b5125 user: btrach tags: trunk, master | |
| 20:38 | (net) don't loop forever when missing +w in ARCAN_STATEPATH check-in: 4e976da227 user: btrach tags: trunk, master | |
| 20:37 | (shmif) don't crash decode when requesting a11y and tesseract unavailable check-in: bb95c658d1 user: btrach tags: trunk, master | |
| 20:36 | (decode) adapt to libvlc 4.0 check-in: 35f30616c5 user: btrach tags: trunk, master | |
| 20:36 | (platform/sdl2) rebuild projection matrix to account for macos reqs check-in: 381aac4493 user: btrach tags: trunk, master | |
| 20:35 | (tests) rename rtmap lua file to match the appl name check-in: 1bc6f1422e user: btrach tags: trunk, master | |
| 08:43 | (core) evict openAL patches from lwa, swap in miniaudio There is still a lot of testing to be done with this one, and even more so for it to be a suitable OpenAL replacement in main. Basic bringup still missing is positional audio and capture devices. check-in: ad30da4287 user: letoram tags: trunk, master | |
| 08:41 | (build) synch vendored sqlite3 check-in: 2dc4ea0d12 user: letoram tags: trunk, master | |
|
2026-06-22
| ||
| 12:51 | (net) only log shmif messages in verbose build check-in: c96230a344 user: letoram tags: trunk, master | |
| 05:15 | (doc) synch contributions to changelog check-in: 172319bc42 user: letoram tags: trunk, master | |
| 05:09 | (terminal/tsm) fix to CSI PPM/CPL check-in: d159262ca2 user: btrach tags: trunk, master | |
| 05:08 | (terminal/tsm) fix formatting of breaks check-in: 97e841b85f user: btrach tags: trunk, master | |
| 05:08 | (core) wrong clipping calculation for tpack raster check-in: 53e731ed68 user: btrach tags: trunk, master | |
| 05:07 | (terminal/tsm) fix screen erase check-in: 83a5cff483 user: btrach tags: trunk, master | |
| 05:07 | (conductor) don't poll empty evsrc_pollset check-in: 7590ec7d1d user: btrach tags: trunk, master | |
| 05:06 | (tui/readline) Correctly choose direction for Up/Down keys in the lower half of canvas. check-in: 9d64dd60f6 user: btrach tags: trunk, master | |