reassemble bug

[charryshi]

Hi,I found a reassemble bug Under specific condition.
Here is the example log

{"timestamp":1630031299.3498211,"station_id":"test-01","channel":1,"freq":127.275,"level":-11,"error":0,"mode":"2","label":"C1","block_id":"M","ack":false,"tail":"B-xxxx","text":"msg example 1-1","end":true,"assstat":"in progress"}
{"timestamp":1630031306.879694,"station_id":"test-01","channel":1,"freq":127.275,"level":-11,"error":0,"mode":"2","label":"C1","block_id":"N","ack":false,"tail":"B-xxxx","text":"msg example 1-2","end":true,"assstat":"in progress"}
{"timestamp":1630031314.1785741,"station_id":"test-01","channel":1,"freq":127.275,"level":-11,"error":0,"mode":"2","label":"C1","block_id":"O","ack":false,"tail":"B-5528","text":"msg example 1-3","end":true,"assstat":"in progress"}
{"timestamp":1630031320.9381371,"station_id":"test-01","channel":1,"freq":127.275,"level":-11,"error":0,"mode":"2","label":"A4","block_id":"P","ack":false,"tail":"B-xxxx","text":"/WUHDCYA.FS1/FSM 0228 210827 ZXXXX\r\nCXA1111 CDA REJECTED\r\nCLEARANCE CANCELLED\r\nREVERT TO VOICE PROCEDURES\r\n1A01","assstat":"skipped"}

not msg lost, another type msg(label A4) started at block_id P after msg block_id O (label C1),but label C1 series not finished at this time, but after block_id P send will be out of sequence from reassemble

{"timestamp":1630031327.961997,"station_id":"test-01","channel":1,"freq":127.275,"level":-11,"error":0,"mode":"2","label":"C1","block_id":"Q","ack":false,"tail":"B-xxxx","text":"msg example 1-4","end":true,"assstat":"out of sequence"}

now
static uint32_t acars_key_hash(void const *key) {
acars_key *k = (acars_key *)key;
uint32_t h = la_hash_string(k->addr, LA_HASH_INIT);
h = la_hash_string(k->label, h);
h = la_hash_string(k->msn, h);
return h;
}

but I can't remove label from the code. I think maybe libacars can add another one identifies to make sure to the address msg is continuously ,not require prev_seq_num is continuously.

thanks a lot.

[szpajder]

Thanks for pointing this out. I am aware of this feature (it's called multiblock nesting), however when I was developing the reassembly code I couldn't find an evidence if it is actually used in practice. Apparently it is. Do you see this situation often?

Anyway, I'll put this feature on my TODO list. Can't give you an ETA right now though - it requires quite a significant rework and thorough testing afterwards.

[charryshi]

Thanks for reply.I saw maybe 10% reassembly error because this question(the others are parity or crc error for receiving).
I think to fix it will need rewrite code structure, I glade to hear it on TODO list.
This project is great, thanks for your working.