Many hyperlinks are disabled.
Use anonymous login
to enable hyperlinks.
50 most recent check-ins
|
2024-12-21
| ||
| 12:46 | (decode) expose wait_for_fail path on av as well Leaf check-in: 60bc04103b user: void tags: master, trunk | |
| 06:37 | (net) internal-queue handler set proper status code check-in: 949366b353 user: void tags: master, trunk | |
| 06:36 | (net) use request streamid throughout check-in: 3eb579d8f7 user: void tags: master, trunk | |
| 05:54 | (net) streamid vs identifier testing check-in: 852e3c83a2 user: void tags: master, trunk | |
| 02:11 | (net) separate handling of bchunkreq for appl, appl-res and priv-res check-in: b5df78939b user: void tags: master, trunk | |
| 02:03 | (net) match reqfail/streamstatus to id forwards to raw caller correctly, but server-end rejects valid check-in: e48dbfd1e5 user: void tags: master, trunk | |
|
2024-12-20
| ||
| 23:56 | (net) also use the right identifier on dir-work-reqfail check-in: 7c7ee261f4 user: void tags: master, trunk | |
| 23:54 | (net) use stream identifier in streamstatus requests check-in: 1baa254ea6 user: void tags: master, trunk | |
| 21:47 | (shmif) add .identifier to BCHUNKSTATE This was necessary for a12 handling of BCHUNK_IN/BCHUNK_OUT and retaining identifier routing versus the raw API. check-in: e4b708794f user: void tags: master, trunk | |
| 14:27 | (net) remove pending_blobs tracking check-in: 69d007f821 user: void tags: master, trunk | |
| 14:26 | (net) switch unlink/alloc blob chain to take root check-in: df7cc0a2a4 user: void tags: master, trunk | |
| 13:29 | (net) rename blob_out to blob_xfer and prepare for blob_in queue check-in: 8ed7690888 user: void tags: master, trunk | |
| 13:25 | (lua) open_nonblock(vid, write = bool) wrong BCHUNK direction check-in: c13e1ea4e9 user: void tags: master, trunk | |
| 13:12 | (a12) dir-worker mirror STREAMSTATUS to ack stream completion check-in: 495fe01764 user: void tags: master, trunk | |
| 13:11 | (net) send STREAMSTATUS(1.0) when queued stream completes check-in: e924d543f8 user: void tags: master, trunk | |
| 01:59 | (net) lifecycle for taggd bstream, log pending-state check-in: 5eab47f21f user: void tags: master, trunk | |
| 01:41 | (net) log more fields for parent resource request check-in: 0975ddd2ec user: void tags: master, trunk | |
| 01:39 | (net) move file ul/dl handling from _dir to a12.c This will break the current implementation, troubleshooting pending. The longer story is that the workarounds used in _dir to handle submitting transfers were hacky and can be cleaner merged inside of the regular a12 handling by matching to the corresponding SHMIF events. What this first round does is to check on enqueue if any special management is needed to account for the other end being a directory. If so, the proper BCHUNKSTATE event is tied to the pending transfer, and submitted before the bstream header gets sent. This should give better queue handling for uploads, but is incomplete for download and download notification still. check-in: 7b61c86944 user: void tags: master, trunk | |
| 01:37 | (lua) drop unused helper, change @stdin to have proper fsrv type check-in: 352f6b7221 user: void tags: master, trunk | |
| 01:35 | (doc) minor note on open_nonblock re-arming check-in: ce596fd354 user: void tags: master, trunk | |
|
2024-12-19
| ||
| 19:07 | (net) truncate in dir_worker on new upload check-in: ff1dfc03db user: void tags: master, trunk | |
| 18:14 | (net) --get-file .priv .index tag@ working check-in: 5a07bb9bca user: void tags: master, trunk | |
| 18:12 | (net) add .priv store resolution for .index check-in: 175f19860a user: void tags: master, trunk | |
| 12:13 | (net) Initialises bool variables to avoid undefined behaviour in clang check-in: a09a8c5410 user: cipharius tags: master, trunk | |
| 01:06 | (distr) synch vendored stb_ check-in: 1612ae47ac user: void tags: master, trunk | |
|
2024-12-18
| ||
| 15:00 | (distr) swap out logo check-in: caef760fee user: void tags: master, trunk | |
|
2024-12-13
| ||
| 21:11 | (doc) start outlining a12 spec check-in: e8968f50c9 user: void tags: master, trunk | |
| 19:23 | (lua) change setglobalint to setglobalnum to avoid accidental num->int thanks @bitcompost check-in: f66307e4c4 user: void tags: master, trunk | |
|
2024-12-12
| ||
| 11:06 | (net) block null-deref on 'no/yes/remember/ query check-in: 83f7862e84 user: void tags: master, trunk | |
| 11:06 | (net) reference the other base64 implementation check-in: 21459e27e3 user: void tags: master, trunk | |
| 11:05 | (build) don't pull in base64.c into arcan-net The implementation actually used comes from a vendored .h that should be changed to a sidechannel robust implementation as it's used to pass keymaterial from dir to worker. check-in: 5d6bb45d70 user: void tags: master, trunk | |
|
2024-12-11
| ||
| 06:16 | (net) Fix pending packets a12 statistics computation check-in: 64d0d8150a user: cipharius tags: master, trunk | |
|
2024-12-07
| ||
| 18:52 | (distr) generalise builtin/mouse.lua lmb/rmbhandler and forward id check-in: 2761a3a3fa user: void tags: master, trunk | |
| 18:06 | (platform) allowlist more XDG_ paths and LASH_BASE check-in: fdbb4288b3 user: void tags: master, trunk | |
| 16:56 | (distr) synch mouse script changes to durden upstream check-in: f5a9820a80 user: void tags: master, trunk | |
|
2024-12-06
| ||
| 11:26 | (net) switch .index to .appl-index in order to use .index for private store This should only affect the communication between srv and worker, as there are protocol commands explicit for requesting the source/appl index of the directory. check-in: 8201f5ee92 user: void tags: master, trunk | |
|
2024-12-02
| ||
| 06:48 | (net) request_parent_resource didn't flush correctly This is a part fix for an ongoing race condition in the initial setup window highlighted by the tests with --put-file. Requesting a file from the private namespace keystore will sometimes not receive the BCHUNK_IN/BCHUNK_OUT when some other inbound event should've been queued in between. The 'continue and queue more' errored out on the wrong return state. check-in: b917240213 user: void tags: master, trunk | |
| 06:42 | (net) cancel stream transfer on EOF Previous binary transfers have all been using fix boundary, so this edge condition wasn't tested. It turned out that it was never marked as completed (read=0) so it just kept waiting. check-in: 69fb262d77 user: void tags: master, trunk | |
|
2024-12-01
| ||
| 01:29 | (net) have worker forward blob requests to parent The missing piece in client to server --put-file should now just be allocation in the keystore allocation for private and the permission check with appl dev- verified content store. check-in: 627ca04764 user: void tags: master, trunk | |
|
2024-11-30
| ||
| 22:38 | (net) error message in @tag vs tag@ check-in: 57ca7aee35 user: void tags: master, trunk | |
| 15:33 | (net) tested appl-download from directory check-in: a341108ddf user: void tags: master, trunk | |
| 14:34 | (net) build fixes to directory server check-in: d94559db6c user: void tags: master, trunk | |
| 14:32 | (net) rework how bchunkreq's behave in directory mode This will take some iteration to get back into the current state and breaks backwards compatibility for directory mode running appls. It is needed to implement how developer pushed appl specific storage is handled, as well as for user-private data management. The change moves the applid:name format from the bstream transfer, and the extid use in bchunrequests. The size field now has an overloaded namespace specifier, and the request part of the binary stream command is used to carry namespace. The namespace identifier corresponds to the applid, with 0 being reserved for private storage. check-in: f1e4f72a5c user: void tags: master, trunk | |
| 08:19 | (net) rework how 'remember' interactive works There are a few nuances to this. By default the trust_domain option ("outbound") is used to remember an unknown key. If the user specifies a custom tag, it would not be triggered when making a new connection as it only explicitly checks the trust-domain. To remedy this, we also register the key in the hostkeys part of the store, so: arcan-net arcan.divergent-desktop.org -> unknown key, do you trust? remember -> dd arcan-net dd@ would work on successive calls. This has the caveat that there is no differentiation for the key, the 'default' one used for outbound will be kept compared to how it works when you manually run: arcan-net keystore dd arcan.divergent-desktop.org which explicitly generates a new outbound key for use with only that tag. check-in: 10bf5a2e86 user: void tags: master, trunk | |
| 08:15 | (shmif) add overloaded form of BCHUNKHINT These are just alternate interpretation of the fields for requesting an input (file-open) when used across the network and we need the option to select a namespace. check-in: 405184f8ac user: void tags: master, trunk | |
|
2024-11-29
| ||
| 13:13 | (net) write out information on unknown beacon senders check-in: 4abb656ae5 user: void tags: master, trunk | |
| 13:13 | (net) scribble some notes on directory mode with bchunk check-in: dfd447adf9 user: void tags: master, trunk | |
| 12:32 | (net) workaround for confused interpretation of STATEDIR in resolving keystore root ARCAN_STATEPATH is used to pass the root for all state store as far as arcan is concerned, and this is used with all frameserver creation. The keystore subsequently reserved the 'a12' subdirectory for use as a keystore as to not clash with anything else stored there. Directory mode using a config- file on the other hand would treat the STATEPATH as being the keystore root as it is set explicitly through the lua. This clashed when trying to beacon broadcast the keystore that is also used to seed the directory. This commit adds a workaround for that by checking for the keydirs in root first, and if they don't exist, follow the 'a12' format. check-in: 2a24e09032 user: void tags: master, trunk | |
| 07:46 | (build) suppress some warnings check-in: cfe794ef39 user: void tags: master, trunk | |
| 01:22 | (net) directory server minor constant cleanup check-in: d71445d982 user: void tags: master, trunk | |