The Linux cacheline probe first tries <linux/cache.h>, but that header is not part of the installed uapi headers on Arch Linux. When the compile probe fails, configure falls back to a host_cpu mapping and currently aborts for riscv64 with:
Unrecognized CPU architecture (riscv64) on Linux fallback path.
Handle riscv* in both cacheline fallback maps and use a 64-byte cacheline. That matches the Linux RISC-V kernel default L1_CACHE_SHIFT value of 6 and the value reported on a riscv64 machine through getconf LEVEL1_DCACHE_LINESIZE and sysfs cache coherency_line_size.
rak::regex was a hand-rolled glob matcher supporting only '*'
wildcards. Replace it with the standard POSIX fnmatch() which supports
the same glob patterns plus '?' and '[...]' character classes.
- src/core/manager.cc: use fnmatch for directory entry filtering
- src/command_download.cc: use fnmatch for file path pattern matching
- Makefile.am: remove rak/regex.h from EXTRA_DIST
- rak/regex.h: deleted, no longer needed
Control::handle_shutdown() closes m_directory_events (the inotify directory
watch) only in the normal-shutdown branch (!m_shutdownQuick). On a quick
shutdown (SIGTERM -> receive_quick_shutdown) the m_shutdownQuick branch skips
it, so the watch stays registered in the poll. ~Control() then destroys the
still-open directory_events and Event::~Event()'s assert(m_poll_event == nullptr)
(added with the 0.16.13 poll/event rework) aborts:
main -> Control::~Control() -> ~directory_events -> ~Event() -> abort
Close it in Control::cleanup() instead, which runs on every shutdown path after
the session is saved; close() is idempotent, so the normal path that already
closed it in handle_shutdown() is unaffected.
Reproducible: configure a directory.watch.added watch, then SIGTERM the client
-> SIGABRT (core in Event::~Event); SIGINT (normal shutdown) is clean. With this
change SIGTERM exits cleanly too.
Since the 0.16.13 callback/poll rework, an EPOLLERR on stdin (controlling
terminal or pty hangup) reaches Poll::process(), which aborts the whole
client with an internal_error because InputEvent never registered for error
events:
Poll::process() received error event for event not in error: input-fd:0
Register stdin for error events (insert_error) and handle event_error() by
dropping stdin from the poll set with this_thread::poll()->remove_and_close().
rtorrent then keeps running without keyboard input instead of dying.
remove() guards on the fd state (is_open()), mirroring SCgiTask: event_error()
clears the fd after remove_and_close(), and the shutdown path (Control::cleanup)
still calls remove(); without the guard the second remove_and_close() throws
'event not found' via event_mask(). insert()/remove() take the thread poll
implicitly (this_thread::poll()) instead of a Poll* argument.
Replace the union-based reinterpret_cast type erasure in command_base
with an alignas char buffer + typed copy/destroy helper pointers.
set_function<T>() placement-news the correct std::function<T> type
at the buffer address, and stores per-type copy/destroy helpers so
that the copy ctor, assignment, and destructor always operate on the
actual type rather than assuming base_function.
_reinterpret_cast<T&> access of t_pod remains zero-overhead and is
now well-defined because the object was constructed at that address
as T via placement new.
Fixes#1818
Replace the raw ~uint32_t sentinel with a named constant
Download::default_resume_flags that masks out the open_enable_fallocate
bit. This prevents flag_fallocate from being set on all files when
open_throw() reads resume_flags() before explicit flags are configured.
The sentinel value (~uint32_t & ~open_enable_fallocate) retains the
full range as a 'not set' marker while being safe to pass through
Download::open() without unintended fallocate.
Use Download::hash_error_message() to get a descriptive error string when
hash check fails due to I/O error, instead of relying on errno which could
be zero.
Also fix Manager::receive_hashing_changed() where set_hash_failed(true) was
called without setting d.message when catching local_error during hash
check (e.g. 'too many open files').
Commit 6488131 ("Fix RPC/SCGI security and crash bugs by @sirus20x6")
replaced std::vector<std::unique_ptr<const char>> storage with
std::vector<std::string> and returned back().c_str() to the xmlrpc-c
registry as the per-method server_info pointer.
This is unsafe for any method name short enough to be SSO-stored
(<= 15 chars on libstdc++): such a string keeps its buffer inside the
std::string object itself. When a later push_back reallocates the
vector and move-constructs the existing elements into a new buffer,
the previously returned c_str() pointers — captured by xmlrpc-c at
registration time — dangle into freed memory.
Because xmlrpc-c does not dereference server_info until a call
dispatches, the failure surfaces later as nondeterministic garbage
in fault strings, e.g.
faultString: Command "thod." does not exist. (load.start, log.xmlrpc, log.execute)
faultString: Command "in_rate" does not exist. (log.add_output)
faultString: Command "" does not exist. (log.open_file)
faultString: Command "+U" does not exist. (method.set_key)
Long-named methods (e.g. system.client_version at 21 chars) are
heap-allocated above the SSO threshold and escape the bug because
the heap buffer's address is preserved across the vector move.
Switch the storage to std::deque<std::string>: per [deque.modifiers]
push_back does not invalidate references to existing elements, so
the std::string objects do not move and the c_str() pointers handed
to xmlrpc-c remain valid for the program's lifetime. The body of
store_command_name is unchanged.
Fixes the use-after-free; preserves the std::string-based storage
the original commit aimed for.
When an SCGI client closes the connection before rtorrent finishes sending
the response, send() in SCgiTask::event_write() returns -1 with errno EPIPE.
EPIPE was grouped with EAGAIN/EINTR as a non-fatal retry-later condition, so
the task was not closed and its descriptor stayed registered for EPOLLOUT. A
broken socket is reported writable immediately, so epoll_wait() returns it on
every iteration and the SCGI thread spins at 100% CPU on one core
indefinitely. The dead connection fd is also leaked (stays ESTAB).
EPIPE is terminal here, not retryable: the peer is gone and the response can
never be delivered. Close the task on EPIPE, matching event_read(), which
already closes on any recv() error other than EAGAIN/EINTR.
Reproduction: open the SCGI socket, send a complete RPC request, then
shutdown(SHUT_RDWR)/close before reading the reply. Stock: the rtorrent-scgi
thread goes to 100% CPU and the connection leaks. With this change: CPU stays
at 0% and the descriptor is closed.
- Add TORRENT_WITHOUT_NCURSES macro in scripts/checks.m4
- Clear CURSES_LIBS/CFLAGS in the macro instead of if/else in configure.ac
- Restore simple LIBS/CFLAGS lines in configure.ac
- Add missing set_escdelay stub
Add a dummy curses stub header that provides all ncurses types, macros,
and no-op function stubs. When configured with --without-ncurses, the
build uses this stub instead of linking to the real ncurses library.
This allows rtorrent to be built for daemon-only usage (e.g. with
ruTorrent or Flood) without requiring ncurses to be installed.
Closes#1613
Jari Sundell