No functional change, since with the parent commit, we no longer treat
"DirRemoteness::local" different from "DirRemoteness::remote", but we might
do so in future, so make sure we don't give a false positive here.
Non-Linux systems have ST_LOCAL or MNT_LOCAL, so no unknowns there.
See #10434
mmap() fails with ENODEV on remote file systems. This means we always fail
to read any old history on network file systems on Linux (except on the file
systems we recognize which are NFS, SMB and CIFS).
Untested, so I'm not sure if this works.
Fixes#10434
We no longer use RAII for enabling/disabling these, so a full object is
overkill. Additionally this object doesn't allow us to recover from the case
where we receive SIGTERM while inside terminal_protocols_{enable,disable}.
We can simply run disable another time since they're idempotent. Untested.
When I run a command with leading space, it is not added to the on-disk
history. However we still call History::save(). After 25 of such calls,
we rewrite the history file (even though nothing was written by us).
This is annoying when diagnosing #10300 where the history of the current
shell (but not other shells) is broken; because the history rewrite will
make the problem go away. Let's not save in this case, to make it easier to
run commands to inspect the state of the history file.
Given a history like
- cmd: echo OLD
when: 1726157160
\x00\x00\x00- cmd: echo leading NUL bytes
when: 1726157160
- cmd: echo NEW
when: 1726157223
offset_of_next_item() happily records 3 items even though the second item
is corrupted.
decode_item() fails which makes the caller stop loading any older items --
we got knee capped.
Avoid this horrible failure mode by skipping over these items already in
offset computation. For now we still lose the corrupted item itself.
In future we should probably try to delete the NUL bytes or avoid the
corruption in the first place.
See #10300 and others.
This should make the sort have a strict weak ordering, which rust
requires since 1.81 (or it will panic).
Note: This changes the order, but that's *fine* since the current
order is random weirdness anyway.
Fixes#10763
Commit b00899179 (Don't indent multi-line quoted strings; do indent inside
(), 2024-04-28) made parse_util_compute_indents() crash on `echo "$()"'x`.
After recursively indenting the command substitution, we indent the "'x
suffix. We skip the quoted part by setting "done=2". Later we wrongly
index "self.indents[done..range.start+offset+1]" (= "self.indents[2..1]").
Fix this by making sure that "start >= done", thus not setting any indents
for the quoted suffix. There is no need to do so; only the first character
in each line needs an indent.
In particular, this fixes the case
ctrl-r foo ctrl-r
where foo substring-matches no more than one page's worth of results.
The second attempt will fall back to subsequence matching which is wrong.
In case a terminal resize[1] causes us
to repaint a multi-line prompt that changes width like
function fish_prompt
for i in 1 2 3
random choice 'aaaaaaaaaaaaaaaaaaaaaaaaaaaaa' 'bbbbbbbbbbb'
end
end
we add a clr_eol after each line[2] , to make sure
that a "b" line does not have leftover "a" letters
(80aaae5b7 (Clear to end of each line in left prompt, 2020-10-25)).
Unfortunately, if a prompt line takes up all the columns, clr_eol will
wrongly clear the last column. Reproduce with
function fish_prompt
string repeat $COLUMNS -
echo "$PWD> "
end
and observe that the last "-" is missing.
Previous (reverted) attempt d3ceba107 (Clear to eol before outputting line
in multi-line prompt, 2021-05-17) found the right fix but had an off-by-one
error which reintroduced the leftover "a" letters in the "random choice"
prompt above.
Given prompt string "aa\nbb\ncc", it wrongly printed
clr_eol "aa" clr_eol "\nbb" "\ncc"
Observe that the first line is cleared twice, while the second line is
never cleared. Fix that.
[1]: or an async "commandline -f repaint" triggered by a uvar change /
async prompt update
[2]: except after the last line where we probably already emit clr_eol
elsewhere..
Alternative fix: emit both clr_eol and clr_bol *before* drawing the current
line. However, if fish and the terminal disagree on character width, that
approach might erase too much.
Closes#8164
I guess it's nice to know that these two are the same but that info is not
needed here, it just adds confusion. The user must have pressed ctrl-j if
we get here, so echo that back.
See the parent commit.
This is just too confusing; \b sounds like it would map backspace but it's
actually just ctrl-h. Backspace is a different key ("bind backspace"),
so let's move away from \b.
Reproduce by typing ctrl-h in fish_key_reader, or, for even more confusion,
use a terminal like tmux and type ctrl-backspace which also sends ctrl-h.
I've thought about changing \b (and its aliases like \ch and \x08) to mean
backspace but that seems like unnecessary breakage, since they all already
mean ctrl-h, and can usually be mapped independent of backspace.
See the discussion in #10738
Make fish-printf no longer depend on the widestring crate, as other clients
won't use it; instead this is an optional feature.
Make format strings a generic type, so that both narrow and wide strings can
serve. This removes a lot of the complexity around converting from narrow to
wide.
Add a README.md to this crate.
This gave a weird error when you did e.g. `math Foo / 6`:
"Missing Operator" and only the "F" marked.
Adding an operator here anywhere won't help, so calling this an
"Unknown function" is closer to the truth. We also get nicer markings
because we know the extent of the identifier.
Fix 7308dbc7a (fish_indent: Prevent overwriting file with identical content,
2024-07-21) in a different way by passing O_TRUNC again.
If we don't want regressions we could use code review.
We used deque in C++ because this vector may be large, and so it avoids
repeated re-allocations. But VecDeque is different in Rust - it's contiguous -
so there's no benefit. Just use Vec.
For implementation reasons, we special-case cd in several ways
1. it gets different completions (handle_as_special_cd)
2. when highlighting, we honor CDPATH
3. we discard autosuggestions from history that don't have valid path arguments
There are some third-party tools like zoxide that redefine cd ("function cd
--wraps ...; ...; end"). We can't support this in general but let's try to
make an effort.
zoxide tries to be a superset of cd, so special case 1 is still
valid but 2 and 3 are not, because zoxide accepts some paths
that cd doesn't accept.
Let's add a hack to detect when "cd" actually means something else by checking
if there is any --wraps argument.
A cleaner solution is definitely possible but more effort.
Closes#10719
In some cases we add the wildcard twice.
$ fish -c '../jj; complete -C"ls cli/*/conf/tem"'
cli/*/*/config/templates.toml
Fix that. Test in the next commit.
There seems to be another bug in 3.7.1 where we fail to apply this completion
to the command line. This appears fixed. (FWIW we might want to revert
the quoting change in completion_apply_to_command_line(), maybe that one
accidentally fix this).
Fixes#10703
The HashMap is used to generate the __fish_describe_command integration
completions. Given the nature of the allocations and the numbers that we use, a
BTreeMap would theoretically perform better. Benchmarks show a 2-9%
improvement in completion times consistently in favor of BTreeMap.
Warnings were appearing under GCC 13.2
(void) alone is insufficient under modern compilers, workaround with logical
negation taken from GCC bug tracker.
Worth including because mold is rather popular in the rust world and because the
bug affects mold versions coincident with the development of the fish rust port.
The bug affects all currently released versions of mold from 2.30.0 (Mar 2024)
onwards under at least FreeBSD (though quite likely other platforms as well).
See https://github.com/rui314/mold/issues/1338 for reference.
The previous control flow logic wasn't sound and would leave the shell in a hung
state when `break` would be encountered.
The behavior is now straightforward, the shell reads until <Enter> or <q> is
pressed, at which point it aborts.
This was based on a misunderstanding.
On musl, 64-bit time_t on 32-bit architectures was introduced in version 1.2.0,
by introducing new symbols. The old symbols still exist, to allow programs compiled against older versions
to keep running on 1.2.0+, preserving ABI-compatibility. (see musl commit 38143339646a4ccce8afe298c34467767c899f51)
Programs compiled against 1.2.0+ will get the new symbols, and will therefore think time_t is 64-bit.
Unfortunately, rust's libc crate uses its own definition of these types, and does not check for musl version.
Currently, it includes the pre-1.2.0 32-bit type.
That means:
- If you run on a 32-bit system like i686
- ... and compile against a C-library other than libc
- ... and pass it a time_t-containing struct like timespec or stat
... you need to arrange for that library to be built against musl <1.2.0.
Or, as https://github.com/ericonr/rust-time64 says:
> Therefore, for "old" 32-bit targets (riscv32 is supposed to default to time64),
> any Rust code that interacts with C code built on musl after 1.2.0,
> using types based on time_t (arguably, the main ones are struct timespec and struct stat) in their interface,
> will be completely miscompiled.
However, while fish runs on i686 and compiles against pcre2, we do not pass pcre2 a time_t.
Our only uses of time_t are confined to interactions with libc, in which case with musl we would simply use the legacy ABI.
I have compiled an i686 fish against musl to confirm and can find no issue.
This reverts commit 55196ee2a0.
This reverts commit 4992f88966.
This reverts commit 46c8ba2c9f.
This reverts commit 3a9b4149da.
This reverts commit 5f9e9cbe74.
This reverts commit 338579b78c.
This reverts commit d19e5508d7.
This reverts commit b64045dc18.
Closes#10634
Add kqueue-based uvar notifier for BSD
Tested under FreeBSD 13.3.
This also works under all versions of macOS, and has some
benefits over the current notifyd choice.
Mutex is used because of the non-mut `notification_fd_became_readable()` `&self`
reference, but contention is not expected.
This restores a hack to trigger a command line repaint when "$fish_color_*" or
"$fish_pager_color_*" changes. These allow the command line to react immediately
to changes in other sessions or web_config.
This was removed in ff62d172e5 but there does not
appear to be a handler which actually redraws these.
Revert of ff62d172e5
The generated assembly is more or less the same and the previously generated
version had been manually verified, but this PR removes the usage of
`MaybeUninit::assume_init()` and replaces it with direct pointer writes.
This should result in no observable change: it continues to pass the functional
tests and benchmarks identically. The safety of the new code has been verified
with Miri.
[0]: https://github.com/mqudsi/fish-yaml-unescape-benchmark
This does not work as-is ("CSI a" is shift-up, not up).
I'm not sure if we want to implement these.
It's not a regression so there is no pressure.
This reverts commit 350598cb99.
byte_to_symbol was broken because it didn't iterate by byte, it
iterated by rust-char, which is a codepoint.
So it failed for everything outside of ascii and, because of a
mistaken bound, ascii chars from 0x21 to 0x2F ("!" to "/" - all the punctuation).
char_to_symbol will print printable codepoints as-is and
others escaped. This is okay - something like `decoded from: +` or
`decoded from: ö` is entirely understandable, there is no need to tell
you that "ö" is \xc3\xb6.
This reverts commit 423e5f6c03.
Array starts at 0, goes up to 27, that's 28 entries... *BUT* we also
need the catch-all entry after, so it's 29.
To be honest there's got to be a better way to write this.
WezTerm supports CSI u but unfortunately, typing single quote on a German
keyboard makes WezTerm send what gets decoded as `shift-'`.
This is bad, so disable it until this is fixed. In future we should maybe
add a runtime option to allow the user to override this decision.
See #10663
The \e\e\[A style is bad but iTerm and putty (alt-left) use it.
The main motivation for this change is to improve fish_key_reader output.
Part of #10663
As of rust 1.78, the Unix stdlib implementation is affected by the same issue:
pub fn sleep(dur: Duration) {
let mut secs = dur.as_secs();
let mut nsecs = dur.subsec_nanos() as _;
// If we're awoken with a signal then the return value will be -1 and
// nanosleep will fill in `ts` with the remaining time.
unsafe {
while secs > 0 || nsecs > 0 {
let mut ts = libc::timespec {
tv_sec: cmp::min(libc::time_t::MAX as u64, secs) as libc::time_t,
tv_nsec: nsecs,
};
secs -= ts.tv_sec as u64;
let ts_ptr = core::ptr::addr_of_mut!(ts);
if libc::nanosleep(ts_ptr, ts_ptr) == -1 {
assert_eq!(os::errno(), libc::EINTR);
secs += ts.tv_sec as u64;
nsecs = ts.tv_nsec;
} else {
nsecs = 0;
}
}
}
}
Note that there is a small behavior change here -- sleep() will continue
after signals; I'm not sure if we want that but it seems harmless?
Part of #10634
iTerm2 deviates from protocol, so back out c3c832761 (Stop using stack for
kitty progressive enhancement, 2024-08-03) in that case.
Note that we use several ways of detecting iTerm2 (ITERM_PROFILE,
TERM_PROGRAM=iTerm.app, ITERM_SESSION_ID).
LC_TERMINAL seems superior because it works over ssh.
This new one should hopefully go away eventually.
Today fish pushes/pops kitty progressive enhancements everytime control is
transfered to/from fish. This constitutes a regression relative to 3.7.1:
$ fish
$ ssh somehost fish
(network disconnect, now we missed our chance to pop from the stack)
$ bash # or some ncurses application etc
(keyboard shortcuts like ctrl-p are broken)
When invoking bash, we pop one entry off the stack but there is another one.
There seems to be a simple solution: don't use the stack but always reset
the current set of flags. Do that since I did not find a strong use case
for using the stack[1] (Note that it was recommended by terminal developers
to use the stack, so I might be wrong).
Note that there is still a regression if the outer shell is bash.
[1]: https://github.com/kovidgoyal/kitty/issues/7603#issuecomment-2256949384Closes#10603
According to the discussion in #2315, we adopt TTY modes for external commands
mainly for "stty". If our child process crashes (or SSH disconnect), we
might get weird modes. Let's ignore the modes in the failure case.
Co-authored-by: Johannes Altmanninger <aclopte@gmail.com>
Part of #10603
Run
printf \Xf6 | wl-copy # ö in ISO-8859-1
LANG=de_DE LC_ALL=$LANG gnome-terminal -- build/fish
and press ctrl-v. The pasted data looks like this:
$ set data (wl-paste -n 2>/dev/null | string collect -N)
$ set -S data
$data: set in local scope, unexported, with 1 elements
$data[1]: |\Xf6|
we pass $data directly to "commandline -i", which is supposed to insert it
into the commandline verbatim. What's actually inserted is "�".
This is because of all of:
1. We never decode "\Xf6 -> ö" in this scenario. Decoding it -- like we do
for non-pasted keyboard input -- would fix the issue.
2. We've switched to using Rust's char, which, for better or worse, disallows
code points that are not valid in Unicode (see b77d1d0e2 (Stop crashing
on invalid Unicode input, 2024-02-27)). This means that we don't simply
store \Xf6 as '\u{00f6}'. Instead we use our PUA encoding trick, making it
\u{f6f6} internally.
3. Finally, b77d1d0e2 renders reserved codepoints (which includes PUA chars)
using the replacement character � (sic). This was deemed more
user-friendly than printing an invalid character (which is probably not
mapped to a glyph). Yet it causes problems here: since we think that
\u{f6f6} is garbage, we try to render the replacement character. Apparently
that one is not defined(?) in ISO-8859-1; we get "�".
Fix this regression by removing the replacement character feature.
In future we should maybe decode pasted input instead. We could do that
lazily in "commandline -i", or eagerly in "set data (wl-paste ...)".
Commit 29f2da8d1 (Toggle terminal protocols lazily, 2024-05-16) made it so
the wildcard expansion in "echo **" (in a large directory tree) can't be
canceled with ctrl-c. Fix this by disabling terminal protocols already at
expansion time (not waiting until execution).
Using
SHELL=$(command -v fish) mc
Midnight Commander will spawn a fish child with
"function fish_prompt;"
"echo \"$PWD\">&%d; fish_prompt_mc; kill -STOP %%self; end\n",
So fish_prompt will SIGSTOP itself using an uncatchable signal.
On ctrl-o, mc will send SIGCONT to give back control to the shell.
Another ctrl-o will be intercepted by mc to put the shell back to sleep.
Since mc wants to intercept at least ctrl-o -- also while fish is in control
-- we can't use the CSI u encoding until mc either understands that, or uses
a different way of passing control between mc and fish.
Let's disable it for now.
Note that mc still uses %self but we've added a feature flag
to disable that. So if you use "set fish_features all"
you'll want to add a " no-remove-percent-self". A patch
to make mc use $fish_pid has been submitted upstream at
https://lists.midnight-commander.org/pipermail/mc-devel/2024-July/011226.html.
Closes#10640
When applying a wildcard, it's important to keep track of the files that have
been visited, to avoid symlink loops. Previously fish used a FileId for the
purpose. However FileId also includes richer information like modification time;
thus if a file is modified during wildcard expansion then fish may believe that
the file is different and visit it twice.
The richer information like modification time is important for atomic file
writes but should be ignored for wildcard expansion; just use the (dev, inode)
pair instead.
This also somewhat reduces our reliance on struct stat, but we still need it for
fstatat which Rust does not expose.
Returns 0 (true) in case an autosuggestion is currently being displayed.
This was first requested in #5000 then again in #10580 after the existing
workaround for this missing functionality was broken as part of a change to the
overall behavior of `commandline` (for the better).
Since we have a mix of both 0-based and 1-based line numbers in the code base,
we can now distinguish between them by type alone. Also stop using 0 as a
placeholder value for "no line number available" except in explicit helper
functions such as `get_lineno_for_display()`.
I've often needed a way to get the last bit of performance out of unwieldy
completions that involve a lot of string processing (apt completions come to
mind, and I ran into it just now with parsing man pages for kldload
completions).
Since many times we are looking for just one exact string in the haystack, an
easy optimization here is to introduce a way for `string match` or `string
replace` to early exit after a specific number of matches (typically one) have
been found.
Depending on the size of the input, this can be a huge boon. For example,
parsing the description from FreeBSD kernel module man pages with
zcat /usr/share/man/man4/zfs.4.gz | string match -m1 '.Nd *'
runs 35% faster with -m1 than without, while processing all files under
/usr/share/man/man4/*.4.gz in a loop (so a mix of files ranging from very short
to moderately long) runs about 10% faster overall with -m1.
Part of #1842
It's like jump-to-matching-bracket, but jumps right before the bracket
I will use it to mimic vi 'ab' and 'ib' text objects in the next commit
Given complicated semantics of jump-till-matching-bracket, an alternative name
could be 'jump-inside-matching-brackets'. But that would make names non-symmetrical.
I'm not sure what is worse.
Part of #1842
Split to:
- jump_and_remember_last_jump. What previously was called jump, now called
jump_and_remember_last_jump
- jump. Only jump, don't remember last jump. Now it's also possible to pass
vector of targets
The commit is pure refactoring, no functional changes are introduced.
The refactoring is needed for the next commits
Command completion descriptions were not being generated from `apropos`. Well,
they were being generated but that was not being correctly used by fish core.
Not sure when this was broken, but there's a possibility it was during the rust
port.
In addition to simply not working, it seems the old code tried to avoid
allocations but String::split_at_mut() allocates a new string (since one
allocation from the global allocator can't be split into two allocations to be
freed separately). Use `String::as_mut_utfstr()` before splitting the &wstr
instead of splitting the &str to actually do this alloc-free.
The "principal" parser is the one and only today; in the future we hope to
have multiple parsers to execute fish script in parallel.
Having a globally accessible "principle" parser is suspicious; now we can
get rid of it.
The "principal" environment stack was the one that was associated with the
"principal" parser and would dispatch changes like to TZ, etc.
This was always very suspicious, as a global; now we can remove it.
Prior to this commit, there was a stack of ReaderDatas, each one has a
reference to a Parser (same Parser in each, for now). However, the current
ReaderData is globally accessible. Because it holds a Parser, effectively
anything can run fish script; this also prevents us from making the Parser
&mut.
Split these up. Create ReaderData, which holds the data portion of the
reader machinery, and then create Reader which holds a ReaderData and a
Parser. Now `reader_current_data()` can only return the data itself; it
cannot execute fish script.
This results in some other nice simplifications.
This is a start on untangling input. Prior to this, a ReaderData and an
Inputter would communicate with each other; this is natural in C++ but
difficult in Rust because the Reader would own an Inputter and therefore
the Inputter could not easily reference the Reader. This was previously
"resolved" via unsafe code.
Fix this by collapsing Inputter into Reader. Now they're the same object!
Migrate Inputter's logic into a trait, so we get some modularity, and then
directly implement the remaining input methods on ReaderData.
Add round options, but I think can also add floor, ceiling, etc. And
the default mode is trunc.
Closes#9117
Co-authored-by: Mahmoud Al-Qudsi <mqudsi@neosmart.net>
The special input functions self-insert, self-insert-not-first, and
and or used to be handled by inputter_t::readch, but they aren't
anymore with `commandline -f`.
I am unsure if these *would* have worked, I can't come up with a use.
So, for now, do nothing instead of panicking.
This would crash if you ran `commandline -f backward-jump`.
The C++ version would read a char (but badly), this doesn't anymore.
So, at least instead of crashing, just do nothing.
This doesn't pull its weight. Block size is not a particularly big
problem,
and this both complicates the code a bit and would arbitrarily cause issues
if a fish script exceeded 65k lines.
This reverts commit edd6533a14.
This doesn't have any effect on the size of the struct (due to alignment
requirements and padding) but reduces the complexity by turning
Block::wants_pop_env into an emergent property dependent on the type rather than
something we have to manually manage.
We only increment it and check if it's non-zero, we never decrement or check the
actual count. As such, change it to a bool and bring the size of `Block` down
from 32 to 24 bytes.
We almost never access any of this and having it stored directly in the `Block`
struct increases its size (reducing how many we can fit in L1 and L2, and
increasing memory copy traffic).
Gets rid of BlockData::None so we can avoid allocating a Box at all when we have
no data (at the cost of yet-another-wrapper-type), which is the usual case.
This has a few advantages,
* We now statically assert that all fields used by a particular block type are
correctly initialized (i.e. you can't assign the function name but forget to
assign its arguments),
* Conversely, we can match directly on `BlockData` and be guaranteed that the
fields we want to access are initialized and present,
* We reduce the number of assertions, effectively "unwrapping" only once based
off the block type instead of each time we try to access a conditional field,
* We reduce the size of the `Block` struct by coalescing fields that cannot
co-exist, bringing it down from 104 bytes to 88 bytes.
It would be nice to make all of `Block` itself an enum, but it currently
requires `Copy` and we take advantage of that to copy it around everywhere.
Putting these fields directly in `Block` directly would mean a lot more memory
traffic just checking block types.
There's no need for two separate block types when one is merely a variant of the
other. This may have been required under C++ but thanks to sum types (rust's
enums) we don't need to do that any more.
If the backgrounded/stopped job was using the tty, sending it SIGCONT first
might cause it to immediately wake and try to use the tty (which fish still has
control over), causing it to immediately stop again after receiving a SIGTTOU.
We are supposed to send SIGHUP first so that when the process resumes it sees
the queued SIGHUP and executes its registered handler!
There's no guarantee that a condition variable is stateful. The docs for
`Condvar::notify_one()` actually say the opposite:
> If there is a blocked thread on this condition variable, then it will be woken
> up from its call to wait or wait_timeout. Calls to notify_one are not buffered
> in any way.
This test was relying on the main loop obtaining the lock and entering the
condition variable sleep before the thread was scheduled and got around to
notifying the condition variable. If this non-deterministic behavior was not
upheld, the test would time out since it would obtain the lock (either before or
after the variable were updated) then call `condvar.wait()` *after* the variable
had been updated and the condvar signalled, but without (atomically or even at
all) checking to see if the desired wake precondition was fulfilled. As the
child thread had already run and the wake notification was NOT buffered, there
was nothing to wake the running thread.
There really wasn't any way to salvage the test as originally written, since the
write to `ctx.val` was not in any way linked to the acquire/release of the mutex
so regardless of whether or not the main thread obtained the mutex and checked
the value precondition before calling `condvar.wait()`, the child thread's write
could have happened after the check but before the wait() call. As such, the
test has been rewritten to use `wait_while()` but then also updated to bail in
case of a timeout instead of hanging indefinitely (since neither the `ctest`
runner nor the `cargo test` harness was timing out; `cargo test` would only
report that the test had exceeded 60 seconds but as long as it was not executed
with `cargo test -- -Z --ensure-time` (which is only available under nightly),
the test would not halt.
If this test were *intentionally* written to test the scenario that was timing
out, it should be written deterministically in such a way that the main loop
did not run until after it was guaranteed that the variable had been updated
(i.e. by looping until val became 5 or waiting for an AtomicBool indicating the
update had completed to be set), but I'm not sure what the benefit in that would
be since the docs actually guarantee the opposite behavior (the notified state
is explicitly not cached/buffered).
If we have fish code written with the assumption that condvar notifications
prior to *any* call to `Condvar::wait()` *are* buffered, then that code should
of course be revisited in light of this.
Commit 8a7c3ce (Don't abandon line after writing control sequences, 2024-04-06)
was broken by 29f2da8 (Toggle terminal protocols lazily, 2024-05-16), fix that.
Fixes#10529
This makes `path basename` a more useful replacement for the stock `basename`
command, which can be used with `-s .ext` to trim `.ext` from the base name.
Previously, this would have required the equivalent of
path change-extension "" (path basename $path)
but now it can be just
path basename -E $path
Mostly replacing std::<type>::MAX with <type>::MAX.
Surprising here is replacing
.expect(format!(...))
with
.unwrap_or_else(|_| panic!(...))
It explains that this is because the "format!" would always be called.
This enabled the profile in fish_setlocale, which caused startup
profile to always be on, so
```fish
fish --profile file -c 'foo'
```
would show the entire startup as well
Hex float parsing may come about through wcstod, for example:
printf "%f" '0x8p2'
should output 32.0.
Currently we use a not-great fork of hexponent. Hexponent has been dormant for
years, and has some issues: doesn't round properly, allocates unnecessarily,
doesn't handle denormals, is more complicated than necessary.
Just rewrite hex float parsing, fixing those problems and getting us off of this
weird fork.
ThreadId is way slower than it should be for the sense that we use it in; it
doesn't cache the id and allocates an Arc internally.
We don't care about the thread id used in crate::threads correlating with any
other thread id the code uses anywhere (not that it does) because it's only used
for our own bookkeeping. Change to something much simpler instead.
Verified that std::sync::OnceLock<T> compiles to the same assembly at the
*access* site as the Option<T> we were using. The additional overhead upon init
is fine. No need for extra Box<T> indirection for IO_THREAD_POOL.
While obtaining an uncontested mutex from the same thread (without reentrance)
is basically ~free, the use of `MainThread<RefCell<T>>` instead of `Mutex<T>`
makes it clear that there is no actual synchronization taking place, hopefully
making the code easier to understand.
The C++ version of this code simply copied the entire uvar table.
Today we take a reference. It's not clear which one is better.
Removal of locale variables like LC_ALL triggers variable change handlers
which call EnvStackImpl::get. This deadlocks because we still hold the lock
to protect the reference to all uvars. Work around this.
Closes#10513
It is short and simple enough to write yourself if you need it and it encourages
bad behavior by a) always returning owned strings, b) always allocating them in
a vector. If/where possible, it is better to a) use &wstr, b) use an iterator.
In rust, it's an anti-pattern to unnecessarily abstract over allocating
operations. Some of the call sites even called split_string(..).into_iter().
This updates is_windows_subsystem_for_linux() to take a WSL version to test for
(any, v1, or v2) and returns the boolean result depending on the system. I've
benchmarked and when running on regular Linux, this is still just as fast as the
previous binary check; it's only when it's WSL that this takes about 20ns
longer to figure out which variant.
Note that older WSLv2 kernels had a `-microsoft-standard` suffix while newer
ones appear to have a `-microsoft-standard-WSL2` suffix, so we make sure to test
for the least common denominator. (It doesn't matter to us, but note that newer
WSLv2 kernels have four dots in the version string!)
WSL workarounds pertaining to the default Windows terminal or executable
behavior of win32 binaries under a WSL shell are extended to WSLv2 while those
specific to oddities in kernel behavior are confined to WSLv1 only. (It
technically wouldn't hurt to extend them to WSLv2 but there's no good reason to
do so, either.)
A common complaint has been the massive amount of directories Windows appends to
$PATH slowing down fish when it attempts to find a non-existent binary (which it
does a lot more often than someone not in the know might think). The typical
workaround suggested is to trim unneeded entries from $PATH, but this a) has
considerable friction, b) breaks resolution of Windows binaries (you can no
longer use `clip.exe`, `cmd.exe`, etc).
This patch introduces a two-PATH workaround. If the cmd we are executing does
not contain a period (i.e. has no extension) it by definition cannot be a
Windows executable. In this case, we skip searching for it in any of the
auto-mounted, auto-PATH-appended directories like `/mnt/c/Windows/` or
`/mnt/c/Program Files`, but we *do* include those directories if what we're
searching for could be a Windows executable. (For now, instead of hard-coding a
list of known Windows executable extensions like .bat, .cmd, .exe, etc, we just
depend on the presence of an extension at all).
e.g. this is what starting up fish prints with logging enabled (that has been
removed):
bypassing 100 dirs for lookup of kill
bypassing 100 dirs for lookup of zoxide
bypassing 100 dirs for lookup of zoxide
bypassing 100 dirs for lookup of fd
not bypassing dirs for lookup of open.exe
not bypassing dirs for lookup of git.exe
This has resulted in a massive speedup of common fish functions, especially
anywhere we internally use or perform the equivalent of `if command -q foo`.
Note that the `is_windows_subsystem_for_linux()` check will need to be patched to
extend this workaround to WSLv2, but I'll do that separately.
Under WSL:
* Benchmark `external_cmds` improves by 10%
* Benchmark `load_completions` improves by an incredible 77%
On this binding we fail to disable CSI u
bind c-t '
begin
set -lx FZF_DEFAULT_OPTS --height 40% --bind=ctrl-z:ignore
eval fzf | while read -l r; echo read $r; end
end
'
because for "fzf", ParseExecutionContext::setup_group() returns early with the
parent process group (which should be fish's own) , hence "wants_terminal"
is false. This seems questionable, I don't think the eval should make a
difference here.
For now, don't touch it; use the more accurate way of detecting whether
a process may read keyboard input. In many of such cases "wants_terminal"
is false, like
echo (echo 1\n2\n3 | fzf)
Fixes#10504
This hot function dominates the flamegraphs for the completions thread, and any
optimizations are worthwhile.
A variety of different approaches were tested and benchmarked against real-world
fish-history file inputs and this is the one that won out across all rustc
target-cpu variations tried.
Benchmarks and code at https://github.com/mqudsi/fish-yaml-unescape-benchmark
We don't forward this variable for storage in any structs, so there's no reason
to go through an Arc instead of returning the `&'static EnvStack` directly.
NB: This particular change was safe, and passes all tests on its own.
We don't forward this variable for storage in any structs, so there's no reason
to go through an Arc instead of returning the `&'static EnvStack` directly.
These have clearer sync/unsync semantics and now ship with rust itself.
They don't paper over any possible cross-thread issues, and we can specifically
choose which we want for the purpose.
`Parser` is a single-threaded `!Send`, `!Sync` type and does not need to use
`Arc` for anything. We were using it because that's all we had for the parser's
`EnvStack`, but though that is *technically* protected internally by a mutex
(shared with global EnvStack), there's nothing to say that other parsers with a
narrower scope/lifetime on other threads will be necessarily using the same
backing mutex.
We can safely marshal the existing `Arc<EnvStack>` we get from
`environment::principal()` into an `Rc<EnvStack>` since the underlying reference
is always valid. To prove this point, we could have PRINCIPAL_STACK be a static
`EnvStack` and have `environment::principal()` use `Arc::from_raw()` to turn
that into an `Arc<EnvStack>`, but there's no need to factorize this process.
By inverting the order of storage, we can use an `OnceCell`/`unsync::Lazy`
inside the Send/Sync `MainThread<T>` and remove the need for a lock altogether.
It's reasonable since this is only checking to see that the history file
contains the expected format and if it's corrupted but we at least got what we
expect to be the correct key/value pairs, then that's all we can do.
Of course the real motivation is to speed up this very hot function in any way
possible!
On the completions and history thread, the parent function
HistoryFileContents::decode_item() is responsible for ~60% of the CPU time, and
extract_prefix_and_unescape_yaml() alone comprising 14% (of the total).
This change removes allocations in the event that the history item is either
fully or partially plain yaml with no escapes to begin with, and brings down the
execution time of this function to only 7% of the total execution time.
The bulk of the remaining time is spent in wcs2string(), which is called
unconditionally and is naturally alloc-heavy.
This allows running `set` without triggering any event handlers.
That is useful, for example, if you want to set a variable in an event
handler for that variable - we could do it, for example, in the
fish_user_path or fish_key_bindings handlers.
This is something the `block` builtin was supposed to be for, but it
never really worked because it only allows suppressing the event for
the duration, they would fire later. See #9030.
Because it is possible to abuse this, we only have a long-option so
that people see what is up.
Commit a583fe723 ("commandline -f foo" to skip queue and execute immediately,
2024-04-08) fixed the execution order of some bindings but was partially
backed out in 5ba21cd29 (Send repaint requests through the input queue again,
2024-04-19) because repainting outside toplevel yields surprising results
(wrong $status etc).
Transient prompts wants to first repaint and then execute some more readline
commands, all within a single binding. This was broken by the second commit
because that one defers the repaint until after the binding has finished.
Work around this problem by deferring input events again while a readline
event was queued. This is closest to the historical behavior.
The implementation feels hacky; we might find odd situations.
For example,
commandline -f repaint end-of-line
set token (commandline -t)
sets the wrong token.
Probably not a very important case. We could throw an error or make it work
by letting "commandline -t" drain the input queue.
That seems too complicated, better change repaints to not use the input queue
(and fake $status etc). Let's try to do that in future.
Closes#10492
[w]open_dir does not pass O_CREAT, so the mode argument to open is never used.
also, O_CREAT | O_DIRECTORY could not be used (portably) to create a directory.
(on POSIX does not specify what should happen, on Linux it is EINVAL.)
rustc 1.80 now complains about features not declared in Cargo.toml and cfg
keys/values not declared by build.rs to protect against typos or misuse (you
think you're using the right condition but you're not). See
rust-lang/cargo#10554 and rust-lang/rust#82450.
(We're not actually using TSAN under CI at this time, but I do want to re-enable
it at some point — especially if we get multithreaded execution going — using
the rust-native TSAN configuration.)
I'll be updating the `rsconf` crate and patching `build.rs` accordingly to also
handle the warnings about unknown cfg values, but tsan is a feature and not a
cfg and these can be dealt with in `Cargo.toml` directly.
We were passing a slice (and not a vec) to `CString::new()`, meaning it would
allocate a new Vec internally to hold the bytes.
Also document that the resulting CString will be silently truncated at the first
interior NUL.
