* Revert "refactor: remove tr_sessionGetConfigDir() (#3506)"
This reverts commit c50da43ae0.
* Revert "fix: do not leak in tr_getWebClientDir() (#3502)"
This reverts commit 5a75e37033.
Add a shim between the handshake code and the rest of the codebase to
improve decoupling so that a followup PR can add handshake unit tests.
The handshake code no longer directly relies on tr_torrent, tr_session,
tr_dht, or tr_peerMgr.
* refactor: remove tr_address_compare()
* refactor: remove tr_address_to_string()
* refactor: remove NUM_TR_AF_INET_TYPES
* refactor: replace tr_sessionGetPublicAddress with tr_session::getPublicAddress()
* refactor: tr_peerIo() takes tr_address by value
* refactor: replace tr_sessionIsAddressBlocked with tr_session::isAddressBlocked()
* refactor: tr_peerMgrAddIncoming now takes tr_address by value
* refactor: replace tr_address_is_valid_for_peers() with tr_address.isValidForPeers()
* refactor: tr_netOpenPeerSocket takes tr_address by value
* refactor: remove tr_generateAllowedSet()
* refactor: setup_sockaddr takes a tr_address by value
* refactor: tr_netBindTCP() takes a tr_address by value
* refactor: tr_dhtAddNode() takes a tr_address by value
* refactor: remove tr_address_from_string()
* refactor: rename tr_address.isValidForPeers() to .isValidPeerAddress()
* refactor: replace tr_address_from_sockaddr_storage() with tr_address::fromSockaddrStorage()
* refactor: minor cleanup to tr_address::readable()
* Fix testing against reference crypto implementation
Build would fail on CI if using cyassl/wolfssl or polarssl/mbedtls as a
backend and testing it against reference openssl implementation due to
missing include directories, but only if bundled (non-system) b64 is
used.
* Iniitalize size arguments to `wc_DhGenerateKeyPair`
Despite it not being mentioned in the documentation [1] and v4.6 release
notes [2], the two size parameters became [3] not purely "out", but
"in/out", to avoid potential buffer overflows.
[1] https://www.wolfssl.com/documentation/manuals/wolfssl/group__Diffie-Hellman.html
[2] https://github.com/wolfSSL/wolfssl/releases/tag/v4.6.0-stable
[3] 4364700c01
* Add basic support for v2 hashes in transmission-show
* Add sha256 for more libraries
* Fix issue with sha256 digest length
* Add sha256 for polarssl
Note: Bumping miniumum PolarSSL version to 1.3 because of
sha2->sha256 name change.
* Add sha256 for CommonCrypto/ccrypto
* Add sha256 for cyassl
* refactor: simplify tr_swarmGetStats()
* refactor: make addStrike() a tr_swarm member function
* refactor: make updateEndgame() a tr_swarm member function
* refactor: limit scope of some constexpr values
* refactor: make tr_swarm.is_endgame private
* refactor: make tr_swarm.isAllSeeds() a member function
* refactor: add tr_peerMsgs.percentDone()
This moves the `progress` and `have` fields from the `tr_peer` parent
class down into BitTorrent peer subclass, since webseeds by definition
are seeds and have everything.
* refactor: use preferred naming for class members
snake_case for variables, camelCase for methods
* refactor: use semantic type tr_torrent_type_t in libtransmission
* refactor: use semantic type tr_torrent_type_t in gtk client
* refactor: use semantic type tr_torrent_type_t in qt client
* Modernize cache.cc: Convert tr_cache to a class
* Modernize cache.cc: Replaced ptrArray with vector
* refactor: Cache now takes a tr_torrents reference
* refactor: add Key type to Cache
* refactor: avoid std::back_inserter
* refactor: add Cache::flushOldest()
This change has been in the back of my head since I made pull request #2984.
Dropping the zero does not change any of the behaviour.
And this way the format is more familiar to users of the unixoid
octal permissions. Those permissions would not be specified with
a leading zero.
If they were specified with a leading zero, it would mean that some
special permission bits (setuid, setgid or sticky) should be cleared.
Thus removing the leading zero in our settings format removes two
cases of possible confusion.
Additionally the permissions are always padded to three digits to avoid
strangely looking permission strings.
We were incorrectly duplicating most domains because the "do we need
to convert before feeding to psl" test incorrectly returned true for
many domains.
* refactor: make peer_atom.blocklisted private
* refactor: make tr_atomAddrStr a member function
* refactor: make atomIsSeed a member function
* refactor: make tr_isAtom a member function
* refactor: make peer_atom constructor
* refactor: make getReconnectIntervalSecs a member function
* refactor: make peer_atom.fromFirst const
* refactor: make tr_peerIoGetAddress() a member function
* refactor: make tr_peerIo.addr a private field
* refactor: make tr_peerIoIsIncoming a member function
* refactor: make tr_peerIoHasBandwidthLeft a member function
* refactor: make tr_peerIoGetPieceSpeed_Bps a member function
* refactor: make tr_peerIoSupportsFEXT() a member function
* refactor: mark tr_peerIo::supportsDHT nodiscard
* refactor: update naming style for tr_peerIo fields
* refactor: make tr_peerIo.is_seed a private field
* fix: ESPIPE "illegal seek" error while seeding
The root cause is createSocket returning `{}` instead of `TR_BAD_SOCKET`
when we have too many peers in use. This defaulted to fd 0, which wound
up closing stdin.
Commit also includes some const-correctness changes made while tracking
the problem down.
* Modernize blocklist.cc to C++
* Modernize blocklist.cc: Renamed BlocklistFile struct and moved IPv4 range struct inside it
Modernize bandwidth.cc: Review notes; Promote container of blocklists in session to vec of unique_ptrs
Charles Kerr