Commit Graph

65 Commits

Author SHA1 Message Date
Arvid Norberg 55460a60bc fix the documentation generation tool to work for flags, and other global constants. Improve documentation 2019-12-19 01:50:22 +01:00
arvidn 2ba0e5ae2c move the file priority vector throught the disk_io_job, to avoid copies 2018-07-08 18:25:06 +02:00
arvidn 267ca40b28 merged RC_1_1 into master 2018-06-02 13:15:13 +02:00
Alden Torres d6cecee8c0 some code cleanup and minor refactor 2018-05-24 23:46:37 +02:00
Arvid Norberg ccdfd7383e move export.hpp into aux_ folder. put assert symbols in libtorrent namespace 2018-05-14 09:09:01 +02:00
arvidn fe612ca167 replace the TORRENT_CFG ABI check with inline namespaces around the symbol that can change ABI based on build configuration options 2018-05-13 19:25:17 +02:00
arvidn f0227451a6 merged RC_1_1 into master 2018-04-13 01:47:41 +02:00
arvidn aaf9304a3b update year in copyright header 2018-04-09 09:04:33 +02:00
arvidn 5413eae900 declare tag types in-line 2018-03-23 11:21:08 +01:00
Alden Torres f0cd6492bd minor lint oriented cleanup and refactor in headers 2018-02-17 00:52:35 +01:00
arvidn 0731200149 block_size is a constant, no need in passing it around as a variable 2018-01-03 08:45:25 -05:00
arvidn 6785046c2f introduce a download_priority_t type, for uniform an type-safe piece- and file priorities 2017-11-02 08:20:21 +01:00
Steven Siloti 0232dc35cd deprecate lock_files (#2344)
File locking interacts poorly with the path to replace a read only file handle
with a read/write one. For performance reasons the old handle is not closed
until the new one is put in place. With file locking this obviously cannot
work because the first handle is holding a lock on the file. As a result,
file_pool::open_file fails in this case. Even if we dropped the reference to
the read only file handle before attempting to re-open it, the open could still
fail if another thread is holding a reference to the read only handle.

There doesn't seem to be a good way around this. The simple solution would be
to always open files in read/write mode, but that has obvious safety downsides.
Any other solution would significantly complicate the file pool code.
Deprecating file locking seems to be the least bad option.
2017-09-17 10:43:27 +02:00
arvidn 55228756e5 make disk job flags type safe 2017-09-10 08:56:52 +02:00
arvidn b03b7417dd build boost from scratch on linux (to get a newer version). upgrade travis to trusty and newer xcode. python binding fixes 2017-08-14 12:22:06 +02:00
arvidn 6fa181ece8 some cleanup of type-safe flags 2017-07-30 18:20:15 -07:00
arvidn 322007bb6a make save_state_flags a strong type 2017-07-30 09:06:41 -07:00
arvidn 78bbd298a5 convert file_open_mode to type safe flags 2017-07-22 23:58:24 -07:00
arvidn 33a2abebfe submit disk jobs in read_piece() 2017-07-18 07:21:07 -07:00
arvidn 4de9f6a75b remove requester parameter to disk read jobs 2017-06-08 12:38:25 +02:00
arvidn 6967d17a42 make move constructors and move assignment operators noexcept 2017-06-01 23:42:15 +02:00
Arvid Norberg 5344761da4 make flags to move_storage a proper enum class (#2043) 2017-06-01 04:15:15 -04:00
arvidn df299fcb00 use boost::variant for disk_io_job in/out parameter, to support holding a proper disk io buffer handle and string 2017-04-17 15:08:27 -04:00
arvidn 480bde9333 change formatting of namespace declarations 2017-04-12 14:26:35 -04:00
Arvid Norberg c35718d322 use unsigned integer for file and storage flags (#1895)
use unsigned integer for file and storage flags
2017-04-09 01:28:46 -04:00
Arvid Norberg 13b16a63a3 use unsigned type for file flags. make file_pool private to disk_io_thread (#1894)
use unsigned type for file flags. make file_pool private to disk_io_thread. move close_file_interval logic to the disk thread. don't expose file in public interface, and make pool_file_status properly public (not part of file_pool and renamed to open_file_status), this is exposed in public interface via torrent_handle::file_status()
2017-04-08 18:24:50 -04:00
arvidn 51c42adc3e steps towars making file_pool private to disk_io_thread 2017-04-07 08:38:35 -04:00
Arvid Norberg ec37436d49 factor out move_storage function to storage_utils.cpp (#1571) 2017-01-17 08:02:44 -05:00
arvidn 4ebc344e48 make disk_io_thread responsible for the storage object. torrents refer to it by index 2017-01-01 13:54:39 -05:00
Arvid Norberg af126fe507 make buffer_allocator_interface an implementation detail of the disk_io_thread (#1463)
make buffer_allocator_interface an implementation detail of the disk_io_thread by using disk_buffer_holder in the disk interface. make chained_buffer use holders instead of free deleter functions and userdata to support putting disk_buffer_holder objects in there. Also make the session buffers be held by holders
2016-12-28 17:47:18 -08:00
arvidn 192ef4962b simplify the buffer_allocator_interface 2016-12-26 08:36:50 -08:00
Arvid Norberg 3f1084d63d strengthened type safety in handling of piece and file indices (#1409)
strengthened type safety in handling of piece and file indices
2016-12-22 10:42:33 -05:00
Arvid Norberg 15ab8f387b cleaning up of disk_interface and improving type safety (#1353)
cleaned up disk_interface and improved type safety. removed unused disk status code disk_check_aborted, and removed status code return value for async_hash
2016-11-26 01:51:47 -05:00
Alden Torres c30027e4a1 more integral type conversion warning fixes, now only needs shorten-64-to-32 (#1349)
more integral type conversion warning fixes, now only needs shorten-64-to-32
2016-11-25 11:17:25 -05:00
arvidn 1d5f42d6d0 convert remaining disk io jobs 2016-11-23 19:02:05 -05:00
arvidn f08501b869 update callback interface for async_check_files and async_flush_piece 2016-11-23 19:02:05 -05:00
arvidn d8662533d5 update callback signature for async_move_storage and async_release_files 2016-11-23 19:02:05 -05:00
arvidn fc9dabe1ab start chipping away at removing disk_io_job from the disk_interface 2016-11-23 19:02:05 -05:00
arvidn 14236993ea collapse logic in piece_manager into storage directly 2016-11-13 02:53:16 -05:00
arvidn ccd539f977 pull check resume data logic out of piece_manager into disk_io_thread and simplify it 2016-11-12 18:03:18 -05:00
arvidn 3ff89f29d7 remove explicit ticking of storage from the torrent. instead make the disk_io_thread do it automatically 2016-11-06 10:26:10 -05:00
arvidn c72053a75c remove unused clear_read_cache function 2016-11-05 21:39:45 -04:00
Arvid Norberg 09cbd0a51d remove deprecated disk_io_thread operations (#1284) 2016-11-02 01:01:27 -04:00
Arvid Norberg b61ebca14a moved async-load-torrent feature out of disk-io-thread (#1282)
moved async-load-torrent feature out of disk-io-thread into a generic worker thread
2016-11-02 01:01:04 -04:00
Alden Torres bb388563b3 refactor to use torrent_info with std::shared_ptr (#1006) 2016-08-17 17:26:35 -04:00
arvidn 8c8c3d5587 replace boost::function with std::function 2016-08-13 07:07:26 -04:00
Arvid Norberg 31b2a03005 some optimizations (#858)
allow disk I/O callback objects to be moved into disk io jobs
2016-06-27 01:02:00 -04:00
Arvid Norberg ed077c8358 use cstdint intead of boost/cstdint.hpp (#833)
use cstdint intead of boost/cstdint.hpp
2016-06-18 14:01:38 -04:00
Arvid Norberg 4d927f4029 modernize disk_buffer_holder to be move-only (#824)
modernize disk_buffer_holder to be move-only
2016-06-16 08:24:41 -04:00
arvidn 96e2693d13 merge RC_1_1 into master 2016-03-19 16:02:46 -04:00