* feat: accept ipv6 string in square brackets for `tr_address::from_string()`
* test: add test case for ipv6 string in square brackets
* fix: include square brackets in host component
According to RFC3986:
host = IP-literal / IPv4address / reg-name
IP-literal = "[" ( IPv6address / IPvFuture ) "]"
* fix: set ipv6-only socket before binding UDP socket
Will return EINVAL on Linux otherwise
* refactor: simplify code using `evutil` when binding TCP socket
* fix: do not set SO_REUSEADDR for listening sockets on Windows systems
Reason: https://stackoverflow.com/a/14388707/11390656
* fix: do not enclose ipv4 address string in square brackets
* fix: `tr_address` should be invalid by default
* fix: allow loopback address for LPD and incoming connections
* fix: `parseCidrline()` should set address type
* code review: keep `memcmp`
* bumping small from 0.2.0 to 0.2.1
bumping libdeflate from 1.18 to 1.19
bumping utfcpp from 3.2.1 to 4.0.5
* rolling back `small` to allow building on NetBSD/amd64.
* keeping submodules on Transmission forks
* bumping `small` to allow building on NetBSD/amd64.
* fix: Sparkle Version Comparator
* Code review: Reducing CFBundleVersion to three components and avoiding versionComparatorForUpdater
* adding +99 when it's a non-beta release
* code review: set CFBUNDLE_VERSION and unset components
* re-adding support for ignoring beta
* refactor: update bep links
* chore: use more appropriate data types
* chore: checkpoint
* refactor: split `can_request()` into each of their own different signature
* chore: checkpoint
* refactor: convert tr_peerMsgsImpl functions to methods
* chore: checkpoint
* refactor: store peer info as reference
* refactor: convert all member variables to private
* chore: re-arrange methods
* refactor: optimise tmp vector default size in `send_ut_pex()`
* chore: housekeeping
* chore: housekeeping
* refactor: avoid `dynamic_cast` when sending cancel
* fix: restore `blocks_sent_to_peer` stat
regression e91af26923
* fix: choke first then reject
* refactor: convert `tr_peerMsgsNew()` to static factory function
* refactor: store `tr_torrent` reference instead of pointer
* Revert "refactor: store peer info as reference"
This reverts commit bb419bf2
---------
Co-authored-by: Charles Kerr <charles@charleskerr.com>
* refactor: add miniupnp 2.3.0 as a submodule
* refactor: remove miniupnpc submodule
* deps: specify the 2.3.0 branch
* build: point to miniupnp monorepo subdirectory miniupnpc
* Fix building miniupnp from Xcode
* fixed miniupnp updateminiupnpcstrings.sh usage
* updating to miniupnpc 2.2.6
* fix build for Windows
* code review: removing the header move
---------
Co-authored-by: Charles Kerr <charles@charleskerr.com>
* chore: fix CGFloat comparison warnings in macOS code.
There are 2 cases:
1. Speed comparisons.
The lowest significant value displayed in UI is 0.1 bytes per sec.
See [NSString stringForSpeed] and [NSString stringForSpeedCompact].
2. Ratio limit comparison.
The lowest significant value displayed in UI is 0.01 (no unit).
This is based on maximumFractionDigits=2 set in related XIB file.
Signed-off-by: Dzmitry Neviadomski <nevack.d@gmail.com>
* chore: fix warning about shadowed variable
CGFloat const difference was used twice in the same scope
Signed-off-by: Dzmitry Neviadomski <nevack.d@gmail.com>
* chore: fix unused block parameter warnings
Signed-off-by: Dzmitry Neviadomski <nevack.d@gmail.com>
* chore: disable GCC_WARN_64_TO_32_BIT_CONVERSION for libtransmission
This warnings are not reported with CMake build.
Signed-off-by: Dzmitry Neviadomski <nevack.d@gmail.com>
* chore: disable CLANG_WARN_STRICT_PROTOTYPES for dht
This is third party target, warning is not enabled with CMake build.
Signed-off-by: Dzmitry Neviadomski <nevack.d@gmail.com>
* chore: disable '-Wexit-time-destructors' warning with CMake.
There are two cases when this is reported in libtransmission:
1. `log_state` in anonymous namespace in `log.cc`.
2. static inline `dh_pool_mutex` in `tr_handshake` in `handshake.h`.
I don't see how this may be improved or how this affects correctness,
so don't nag about that.
Signed-off-by: Dzmitry Neviadomski <nevack.d@gmail.com>
---------
Signed-off-by: Dzmitry Neviadomski <nevack.d@gmail.com>
Version 3.14 was recently released with security fixes, and GitHub
runners have been updated to use it. Don't hardcode the version in
Program Files directory name to support any 3.x. Since 3.x support will
be dropped within a year, need to plan migration to 5.x when it's
released (4.x isn't suitable because of missing heat replacement).
* Support building Qt using CMake 3.28+
CMake 3.28 dropped support for deriving installation prefix(es) based on
PATH environment variable on Windows. Since all the other built packages
already pass necessary prefix path(s) explicitly via CMAKE_PREFIX_PATH
and it works for them, do the same for Qt 6 as well. Building Qt 5 isn't
affected as it doesn't use CMake.
* Fix environment variable name in `windows` build job
Using wrong name resulted in empty `CMAKE_PREFIX_PATH` CMake setting
value, which showed the effects of CMake 3.28 changes yet again.
* refactor: use fmt::print in log.cc
https://github.com/fmtlib/fmt/issues/428#issuecomment-395442159
> You can use fmt::print("...\n") on Windows as well.
Use this mechanism instead of tr_sys_file_write_line()
* refactor: use FILE* in daemon
* refactor: remove unused tr_sys_file_flush_possible()
* refactor: remove unused tr_sys_file_write_line()
* refactor: remove unused tr_sys_file_get_std()
* refactor: remove unused tr_std_sys_file_t
Yat Ho