Previously we would try to send three random TCP relays that we're
connected to to each friend once every 5 minutes. The problem with
this method is that it could take an extraordinarily long time
to share every relay; some relays might be consistently skipped
while others might be sent repeatedly. Moreover, there's no
guarantee that the nodes you try to send are actually online.
This leads to a prety unreliable and flaky way of sharing.
Now we reduce the timer to two minutes, and cycle through the list
trying 3 nodes each share attempt. This guarantees that every online
node in our list gets shared with every friend after a fixed amount of
time (which depends on how many nodes are in the list)
This fixes a bug where file transfers would break when the sendto()
system call failed with errno 11 (usually indicating a full packet
buffer).
The max loops constant has also been greatly reduced, as it was
excessively CPU intensive, and most connections would not benefit
from such a high value. In the future it will be set dynamically.
- Make sender send more data per iteration.
- Make receiver iterate more often while receiving.
Before this commit tox would send at maximum around 4MiB/s. With this
patch sustained speeds of up to 100MiB/s were observed on a
low-latency, high-bandwidth network.
As a consequence of iterating more frequently the receiver's CPU usage
is increased for the duration of the transfer. The data structures
used to represent friends and file transfers cause the sender code use
costly loops that do little real work. This patch makes this problem
more visible: the sender uses more CPU while sending.
Poor network conditions were simulated using the netem kernel
facility: $ tc qdisc add dev lo root netem delay 100ms 50ms \
loss 1% duplicate 1% corrupt 1% reorder 25% 50%
and no adverse behavior was encountered. Tests were conducted
using toxic using both UDP and TCP.
One of these was creating a single 262144 byte stack frame. We now have
a way to check and limit the allocation size of a VLA. The `Cmp_Data`
ones were also fairly large. Now, no allocation is larger than 2KiB
(though rtp.c allocates close to that much).
We enable them on CI, but a default build will compile without them so a
default build doesn't end up with lots of warnings (which we also
disable on CI).
tox_unpack is for unpacking Tox types, and bin_unpack is for
unpacking ints and binary data. This prevents us from creating
dependency cycles due to tox_unpack depending on tox.h
This commit adds functionality for clients to interact with
the DHT, sending getnodes requests to their peers and receiving
nodes in getnodes responses.
These help static analysis and ubsan. We should eventually have all
functions annotated like this with a cimple check to make sure every
pointer has an explicit nullability annotation. The `nullable`
annotation does nothing in GCC, but will be used by cimple to validate
that every parameter has defined nullability.
