Commit Graph
8 Commits
Author SHA1 Message Date
Andy Huska cb0746f3d0 Fixed socketcand dropping valid messaages with 0 data bytes 2022-10-17 10:39:51 -05:00
Andy Huska 3c07785040 Buffer for partially received and/or leftover frame fragments
now working with multiple active busses.

Removed commented out test code and put a fix in for for
the filters list not having its capacity properly reserved
in recalcOverwrite so multiple entries of the same message
show up in the filtered view
2022-10-03 17:02:13 -05:00
Andy Huska e52a74ac7a Found a bug where complete frames being removed from the buffer were being shorted by 1 character
Committing now with debug comments for future reference
2022-10-03 16:36:30 -05:00
Andy Huska 08953a4004 Improved socketcand input buffer handling to reduce lost data.
No longer tossing partial frames, saving the data to concat with future data.
On startup there can be some data loss, but after it gets rolling there is none.
Added rough provision to ensure the buffer doesn't get filled with bad data, but
the decodeFrames recursive calls do a pretty good job of getting rid of it so I
haven't seen the buffer grow after millions of frames coming over a UDP-based VPN.
2022-10-03 11:26:28 -05:00
Andy Huska 60520b5939 Added handing of ok message from RAW mode switch coming in out of order with other messages...
For now I made it its own else statement so it can be detected
2022-09-17 12:52:03 -05:00
Andy Huska 47d07b8bf2 Fixed bug where switching to RAW mode would fail due to the "ok" response being received in the same message as the first data frame
Sometimes when opening socketcand 1 or more busses won't start right and there's a ton of debug data saying "received datagramm: ...."
It seems to happen because the switch to RAW mode is in progress and we receive the OK with a frame in the same datagramm and we don't parse the OK out of the string
but rather expect the whole string to just be OK.  Offending code is in socketcand.cpp SocketCANd::procRXData line 386

Fixed by looking to see if the < ok > message was at index 0 of the incoming message...could also look for it elsewhere (as in index != -1), but not sure if necessary
2022-09-17 12:48:48 -05:00
Andy Huska 7b966fe4a6 Fixed incorrect handling of partial frame tagged onto the end of a complete frame.
Not sure if my method drops the partial or just sends the buffer back to continue filling.
2022-09-16 15:38:54 -05:00
Bertrand c49783a55b adding socketcand support 2021-02-23 09:38:19 +01:00