summaryrefslogtreecommitdiffstats
path: root/src
Commit message (Collapse)AuthorAgeFilesLines
...
* bftw: Fix unbuffered depth-first searchesTavian Barnes2023-10-121-15/+41
| | | | | | | | | | | | | | | bftw() implements depth-first search by appending files to a batch, then prepending the batch to the queue. When we switched to separate file/ directory queues, this was only implemented for the file queue. Unbuffered searches don't use the file queue, so they were all breadth- first in practice. This meant that iterative deepening (-S ids) was actually "iterative deepening *breadth*-first search," a horrible strategy with no advantage over regular breadth-first search. Now it performs iterative deepening *depth*-first search, which at least limits its memory consumption. Fixes: c1b16b49988ecff17ae30978ea14798d95b80018
* bftw: Let iterative deepening work depth-first when sortingTavian Barnes2023-10-121-1/+1
|
* bar: Replace volatile sig_atomic_t with actual atomicsTavian Barnes2023-10-061-8/+12
|
* alloc: Test allocation size overflowsTavian Barnes2023-10-061-2/+2
|
* bfstd: Add a thread-safe wrapper for strerror()Tavian Barnes2023-10-058-7/+63
|
* thread: Define thread_localTavian Barnes2023-10-052-0/+16
|
* config: Fold !__FreeBSD__ into BFS_USE_SYS_CAPABILITY_HTavian Barnes2023-10-052-2/+2
|
* dstring: Limit the special dchar typedef to lint buildsTavian Barnes2023-10-041-7/+2
| | | | | Clang still thinks that alignof(dstr[1]) == 2, so out of an abundance of caution, don't mess with dchar alignment in normal builds.
* thread: s/call_once/invoke_once/Tavian Barnes2023-10-033-3/+3
| | | | call_once() is a reserved identifier from C11.
* ioq: Use io_uringTavian Barnes2023-10-022-42/+288
| | | | Closes #65.
* Formatting fixesTavian Barnes2023-09-2721-99/+93
|
* bit: Fix UINTPTR_WIDTH typoTavian Barnes2023-09-261-1/+1
|
* bit: Use predefined __*_WIDTH__ macros if we canTavian Barnes2023-09-261-24/+69
|
* dstring: New dchar typedef for dynamic stringsTavian Barnes2023-09-2611-64/+85
|
* Use the new list macrosTavian Barnes2023-09-255-26/+22
|
* list: New for_[s]list() macrosTavian Barnes2023-09-251-2/+60
|
* list: Unify formattingTavian Barnes2023-09-251-16/+32
|
* list: New [S]LIST_EMPTY() macrosTavian Barnes2023-09-251-0/+18
|
* list: New [S]LIST_ITEM_INIT() macrosTavian Barnes2023-09-251-22/+52
|
* list: Use (void)(...) rather than do { ... } while (0)Tavian Barnes2023-09-251-31/+30
| | | | This makes everything usable in expression contexts.
* bftw: Share the bftw_state between iterations of ids/edsTavian Barnes2023-09-131-72/+71
|
* parse: Give more ephemeral_fds to -no{user,group}Tavian Barnes2023-09-121-7/+9
| | | | | | | | Fewer than 3 can lead to Assertion failed: (retval->write_queue != -1), function __open_cached_connection, file /usr/src/lib/libc/net/nscachedcli.c, line 224. on a FreeBSD system with LDAP accounts.
* Release 3.0.23.0.2Tavian Barnes2023-09-061-1/+1
|
* bfstd: Work around a FreeBSD-specific msan issueTavian Barnes2023-09-061-4/+23
| | | | Link: https://github.com/llvm/llvm-project/issues/65532
* bfstd: Skip a whole loop in dollar_quote() if possibleTavian Barnes2023-09-061-2/+1
|
* bfstd: Fix printable_len() off-by-oneTavian Barnes2023-09-061-2/+2
| | | | | | | If xmbrtowc() fails, or if xiswprint() is false, then we shouldn't include that wide char in the printable length. Fixes: 19c96abe0a1ee56cf206fd5e87defb1fd3e0daa5
* bfstd: Fix an OOB string index in xmbrtowc()Tavian Barnes2023-09-061-6/+2
| | | | | | | | | | | | | | | | | | | | | | This bug could be reproduced with something like $ bfs -samefile $'\xFA\xFA' bfs: error: bfs: dstrnescat@src/dstring.c:252: wordesc() result truncated or worse, with -DNDEBUG, $ bfs -samefile $'.....................\xFA\xFA' bfs: error: bfs -samefile $'.....................\xFA\xFA\x00\x55\x53\x45\x52\x3D\x74\x61\x76\x69\x61\x6E\x61\x74\x6F\x72 bfs: error: ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ bfs: error: No such file or directory. which prints the memory after the end of the string (in this case, the environment variable USER=tavianator). The bug was caused by the line `*i += len`, which was intended to be `*i = len`. But actually, the right behaviour seems to be `*i += 1`. Fixes: 19c96abe0a1ee56cf206fd5e87defb1fd3e0daa5
* bftw: Enforce the dirlimit strictlyTavian Barnes2023-09-061-19/+17
| | | | | | The previous accounting didn't fully control the number of allocated bfs_dirs, as the dirlimit was incremented once we popped the directory, not when we freed it.
* pwcache: Don't use _SC_GET{PW,GR}_R_SIZE_MAXTavian Barnes2023-08-311-23/+9
| | | | | They tend be 1024, which is a lot of memory per user/group. 128 is usually enough, so start there instead.
* pwcache: Arena-allocate struct passwd/groupTavian Barnes2023-08-311-54/+80
|
* alloc: New varena_grow() functionTavian Barnes2023-08-312-0/+29
|
* trie: New trie_clear() functionTavian Barnes2023-08-313-2/+14
|
* alloc: New [v]arena_clear() functionsTavian Barnes2023-08-312-1/+25
|
* Work around https://github.com/llvm/llvm-project/issues/64946Tavian Barnes2023-08-231-0/+3
|
* ioq: New ioq_slot_monitor() helperTavian Barnes2023-08-231-4/+8
|
* parse: Allow -files0-from an empty set of pathsTavian Barnes2023-08-141-10/+3
| | | | This follows a behaviour change in GNU findutils 4.9.0.
* bfstd: Don't label a declarationTavian Barnes2023-08-081-1/+1
|
* stat: Don't check for statx() mask bits we already checked forTavian Barnes2023-08-071-31/+15
|
* bfstd: Check multiple chars at once for isascii()Tavian Barnes2023-08-071-3/+20
|
* bfstd: Speed up wordesc() by caching isprint()/isspace()Tavian Barnes2023-08-071-2/+29
|
* bit: Add a cast to squelch -Wsign-compareTavian Barnes2023-07-241-1/+1
| | | | I see this only with musl-gcc for some reason.
* Release 3.0.13.0.1Tavian Barnes2023-07-181-1/+1
|
* bftw: Use bftw_file->next for multiple listsTavian Barnes2023-07-181-29/+21
| | | | | A file can be on the to_open and to_read lists at the same time, but otherwise only one list, so we can save memory by sharing the pointers.
* bftw: Use a larger ioq depthTavian Barnes2023-07-181-22/+12
| | | | | | | Now that the dirlimit provides backpressure on the number of open directories, we can use a uniformly larger queue depth for increased performance. The current parameters were tuned with a small grid search on my workstation.
* bftw: Add a queue of directories to unwrapTavian Barnes2023-07-181-7/+22
| | | | | | | | For !BFS_USE_UNWRAPDIR, if a file is still pinned in bftw_closedir(), it has to stay open until its pincount drops to zero. Since this happens in bftw_ioq_pop(), we can't immediately call bftw_unwrapdir() as that adds to the ioq. Instead, add it to a list that gets drained by the next bftw_gc().
* bftw: Add dirs to the end of the queue in bftw_ioq_pop()Tavian Barnes2023-07-181-11/+25
| | | | | | I tried this before in #105 but it led to performance regressions. The key to avoiding those regressions is to put some backpressure on how many bfs_dir's can be allocated simultaneously.
* bftw: Use separate queues for open and closed directoriesTavian Barnes2023-07-171-89/+115
|
* bftw: Check that file->fd == bfs_dirfd(file->dir) earlierTavian Barnes2023-07-171-2/+3
| | | | | | | This has the potential to fail on at least one known platform: macports with the legacysupport implementation of fdopendir(). Link: https://github.com/macports/macports-ports/pull/19047#issuecomment-1636059809
* bftw: Reserve space in the cache before opening filesTavian Barnes2023-07-171-3/+15
| | | | | | | This fixes a storm of EMFILE retries observed with -j1 on a very large directory tree. Fixes: 7888fbababd22190e9f919fc272957426a27969e
* bftw: Pass the whole bftw_state to bftw_openat()Tavian Barnes2023-07-171-510/+451
| | | | | This required shuffling a lot of code around. Hopefully the new order makes more sense.
generated by cgit v1.2.3 (git 2.39.1) at 2024-12-26 16:17:53 +0000