Age | Commit message (Collapse) | Author |
|
before we are sure we must
|
|
|
|
|
|
|
|
first remote
|
|
|
|
|
|
|
|
confuse remotes before and after reconfiguration"
This reverts commit 4475652054616ded16485341b081abf42cc6c08e.
|
|
remotes before and after reconfiguration
|
|
|
|
|
|
Not a single implementation was using them...
|
|
invalid aliasing (hopefully)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
As libsodium has some strange include files (like a version.h), we try to use
absolute include paths whenever possible in fastd now and rename our generated
headers.
|
|
|
|
|
|
|
|
The monotone timestamp is near zero on linux systems, confusing fastd.
|
|
The handshakes are now schedules in a doubly-linked list that is maintained as a
part of the peer structure.
|
|
By using a global keepalive timer, the O(n) keepalive queue purge operation on
every send operation is avoided.
|
|
handshake finish
This ensures that the handshake is repeated in case the finish packet is lost.
|
|
|
|
Everything clang and GCC warn about, except GCC's missing-field-initializers
which are just stupid as they don't allow {} syntax to zero a field.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|