When Transmission listens on both IPv4 and IPv6 interfaces IPv4 listen address
is always passed to CURL's CURLOPT_INTERFACE. In general it stays unnoticed but
if user has multiple IPv6 addresses configured on his system random (first?)
IPv6 address is used. It happens because passed value to CURLOPT_INTERFACE
is not correct -- IPv6 expected but IPv4 is passed.
* refactor: use std::array for keys in transmission-remote
* refactor: use std::array for wide_module_path in platform.cc
* refactor: use std::array for value in tr_env_get_int()
* refactor: use std::array for SysStoreNames in web.cc
* chore: remove unused macro TR_N_ELEMENTS
* fixup! refactor: tr_webseed simplification (#2613)
call connection_limiter.taskStarted in the right place
* perf: put web thread to sleep when inactive
* refactor: reduce nested conditional webseed logic
* refactor: use tr_block_info::Location in webseed
* request larger chunks at a time from webseeds
* fix: CURLOPT_MAXREDIRS to safeguard against loops
Discussion at https://trac.transmissionbt.com/ticket/6110 .
Reported by @x190 and @cfpp2p a few years back.
* fix: make variable a pointer-to-const
* fix: use init-statement to reduce variable scope
* fix: implicit conversion from long to int
* fix: refactor to not nest more than 3 if|for|do|while|switch statements
* fix: make tr_session::setSocketTos() const
* fix: use array.prototype.some instead of a for loop
* refactor: use nullptr instead of NULL
* fix: oops
* refactor: add fetch options object to tr_webRun()
Fold `tr_webRun()`, `tr_webRunWithCookies()`, and `tr_webRunWebseed()`
into a single API that takes an options argument that callers can
customize to their needs.
Also does a surface cleanup pass to the tr_webseed and tr_web internal
structures, e.g. making fields const where possible, not using raw
pointers, and making some fields private.
This revision is still full of code smells. Refactoring the entire
system is overwhelming, so instead I'm doing it in incremental steps.
* refactor: remove effective-url caching in webseeds
The upcoming CURLSH refactor is a more effective way of doing this and
the current appoach makes tr_web harder to refactor, so remove it.
* refactor: webseed rate limiting
The current code that limits the number of parallel fetches is overly
complicated and is also interwoven into the tr_webseed class. Extract
it into a new private helper class "ConnectionLimiter" with a simpler
public API for tr_webseed to use.
* refactor: include <memory> when using shared_ptr or unique_ptr
* refactor: include <cstdio> iff we use it
* refactor: include <cstring> iff we use it
* refactor: include <cstdlib> iff we use it
* refactor: include <string_view> or <string> iff we use it
* refactor: include <array> iff we use it
* refactor: include <ctime> iff we use it
* refactor: include <cctype> iff we use it
* refactor: misc #include cleanups in libtransmission
* perf: don't update queue positions during shutdown
When a torrent is being removed, we try to update the queue positions of
the other torrents. But it's (slightly) expensive and is irrelevant when
the session is closing, so skip it in that case.
* perf: torrentLoadResume shouldn't touch tor->dirty
Loading torrent settings from disk does call a lot of tr_torrentSetFoo()
functions, but since these are the saved settings, they shouldnt' affect
the torrent's 'is dirty' flag.
