Prior to this fix, when completing a command that doesn't have a /, we
would prepend each component of PATH and then expand the whole thing. So
any special characters in the PATH would be interpreted when performing
tab completion.
With this fix, we pull the PATH resolution out of complete.cpp and
migrate it to expand.cpp. This unifies nicely with the CDPATH resolution
already in that file. This requires introducing a new expand flag
EXPAND_SPECIAL_FOR_COMMAND, which is analogous to EXPAND_SPECIAL_CD
(which is renamed to EXPAND_SPECIAL_FOR_CD). This flag tells expand to
resolve paths against PATH instead of the working directory.
Fixes#952
Modify `fish_indent` to emit redirections without a space before the target of
the redirection; e.g., "2>&1" rather than "2>& 1" as the former is clearer to
humans.
Fixes#2899
The relevant standards allow the mbtowc/mbrtowc functions to reject
non-ASCII characters (i.e., chars with the high bit set) when the locale
is C or POSIX. The BSD libraries (e.g., on OS X) don't do this but
the GNU libraries (e.g., on Linux) do. Like most programs we need the
C/POSIX locales to allow arbitrary bytes. So explicitly check if we're
in a single-byte locale (which would also include ISO-8859 variants)
and simply pass-thru the chars without encoding or decoding.
Fixes#2802.
I noticed while fixing issue #2702 that the fish program being tested
was sourcing config.fish files outside of the current build. This also
happens when Travis CI runs the tests but isn't an issue there because
of how Travis is configured to execute the tests.
I also noticed that running `make test` was polluting my personal fish
history; which will become a bigger problem if and when the fishd universal
var file is moved from $XDG_CONFIG_HOME to $XDG_DATA_HOME.
This change makes it possible for an individual to run the tests on
their local machine secure in the knowledge that only the config.fish and
related files from their git repository will be used and doing so won't
pollute their personal fish history.
Resolves#469
pcre2_substitute() now sets the output buffer length to PCRE2_UNSET (~0)
if the output buffer is determined to be too small. This change keeps
track of the buffer size separately where pcre2 can't touch it.
A better fix would be to let pcre2 tell fish what size buffer it needs.
This can be done with PCRE2_SUBSTITUTE_OVERFLOW_LENGTH, but this
requires pcre2 10.21 or later (released January 12), which may be too
new to introduce as a dependency at this point.
Fixes#2743
I've run this more than twenty times through Travis CI (by adding/removing
a comment line). Without this tweak the longest sequence seems to be
around six successful runs.
Expand globs to zero arguments (nullglob) only for set, for and count.
The warning about failing globs, and setting the accompanying $status,
now happens regardless of mode, interactive or not.
It is assumed that the above commands are the common cases where
nullglob behaviour is desirable.
More importantly, doing this with `set` is a real feature enabler,
since the resulting empty array can be passed on to any command.
The previous behaviour was actually all nullglob (since commit
cab115c8b9), but this was undocumented;
the failglob warning was still printed in interactive mode,
and the documentation was bragging about failglob behaviour.
This changes the default escape timeout for the default keybindings (emacs
mode) to 300ms and the default for vi keybindings to 10ms.
I couldn't resist fixing a few nits in the fish_vi_key_bindings.fish file
since I was touching it to set the escape timeout.
My previous commit failed in the travis-ci environment despite passing on my
local computer. This appears to be due to expect timing out looking for the
expected input. See if increasing the expect timeout slightly fixes the
problem.
Introduce a "fish_escape_delay_ms" variable to allow the user to configure the
delay used when seeing a bare escape before assuming no other characters will
be received that might match a bound character sequence. This is primarily
useful for vi mode so that a bare escape character (i.e., keystroke) can
switch to vi "insert" to "normal" mode in a timely fashion.
This fails on e.g. an abbr that uses `env a=b`, like the included test demonstrates.
Unfortunately it decreases the speed again (2s vs 2.2s vs 4s original),
but correctness is more important.
The fix for #2075 inadvertently started unescaping the strings emitted
from `commandline -b`. Only strings emitted with the `-o` flag are
supposed to be unescaped.
Fixes#2210.
Previously 'set -ql' would only look for variables in the
immediate local scope. This was not very useful. It's also
arguably surprising, since a 'set -l' in a function, followed
by a 'set -ql' in a child block, would fail. There was also no
way to check for a function-scoped variable, since omitting the
scope would also pull in global variables.
We could revisit this and introduce an explicit function scope.
Fixes#2502
This isn't pretty, but it fails for, as far as I can see, no _real_
reason.
It doesn't seem to be possible to trigger the failure in real usage, no
matter how fast you press the ESC key followed by something else.
So now this is known and constant travis emails don't help it in any way.
When performing wildcard expansion with a literal path segment,
instead of enumerating the files in the directory, simply apply the
path segment as if we found the directory and continue on. This
enables us to expand strings that contain unreadable directory
components (common with $HOME) and also improves performance, since
we don't waste time enumerating directories unnecessarily. Adds
a test too.
Fixes#2099
Prior to this fix, if you exported a variable in one scope
and then unexported it in the next, it would remain exported.
Example:
set -gx VAR 1
function foo; set -l VAR; env; end
foo
Here 'VAR' would be exported to 'env' because we failed to
notice that the env var is shadowed by an unexported variable.
This occurred at env var computation time, not in env_set!
Fixes#2132
Before this fix, `function -a arg1 name1` would produce a
function named 'arg1'. After this fix, it will produce a
function named 'name'. See #2068 for more.
Before running a command, we add the command to history, so
that if the command causes us to exit it's still captured in
history. But that command should not be considered part of
history when expanding the history within the command itself.
For example, `echo $history[1]` should be the previously
run command, not `echo $history[1]` itself.
Fixes#2028
GNU and BSD `mktemp` handle options differently, and it's a useful
utility for tests. As such, define a common `mktemp` function wrapper
for the test suite.
It might actually be nice to expand this for more flags and support it
globally, but that may result in confusion for any users of BSD mktemp
that expect to be running /bin/mktemp.
Update the test runners so they set up their own environment in
test_util.fish. This simplifies the Makefile and paves the way for
adding utility functions for use in the tests themselves.
Prior to this change, inherited environment variables
would be split on colons, becoming an array. This change
eliminates that behavior. Now environment variables are
always split on the record separator character (ASCII 0x1e),
with the exception of a short whitelist of PATH, MANPATH,
CDPATH. Likewise, exported variables are also exported
delimited by rs, with the exception of the above whitelist.
Fixes#1374, also see #1656
We can't use $PATHS to test the :-splitting because the global config
file adds extra paths based on /etc/paths and /etc/paths.d.
Ideally fish would have a way to suppress behavior like that, but for
the time being it doesn't.
--inherit-variable takes a variable name and snapshots its current
value. When the function is executed, it will have a local variable with
this value already defined. Printing the function source will include
synthesized `set -l` lines for the values.
This is primarily useful for functions that are created on the fly, such
as in `psub`.
Instead of globally marking the state as "in block" when evaluating
blocks/functions, update the "in block" status when pushing/popping
blocks on the parser stack.
Fixes#1729.
On a side note, `status -b` is actually pretty useless, because it
always returns 0 inside of a function (even without this patch).
It seems expect prioritizes the first pattern in the list, instead of
the pattern that matches earliest in the buffer. That seems pretty
stupid, but let's try moving the prompt pattern to the end and see if
that fixes the Travis failures.
Also tweak colored output to reset before the newline instead of after,
so travis behaves better (for some reason reset causes travis to display
the line in black).
Split test_interactive off from test_fishscript and add a new target
test_high_level that tests both.
Add some Makefile magic so the tests can be run serially without using
sub-make, which gets rid of a little noise from the make output.
Rewrite interactive tests to look better.
U+F8FF is the last character in the private use area, but it's also the
codepoint used for the Apple symbol (), which is typeable on US
keyboards in OS X, and so should actually work.
The `--null` flag to `read` makes it split incoming lines on NUL instead
of newlines. This is intended for processing the output of a command
that uses NUL separators (such as `find -print0`).
Fixes#1694.
Directories are completed like commands, because of implicit cd.
However, directories found inside $PATH entries should not be completed,
as implicit cd doesn't work there. Similarly, directories should not be
completed after the `command` builtin.
Fixes#1695.
`exec` removes fish from the shell "stack", so SHLVL needs to be
decremented to match. This means `exec fish` will result in the same
SHLVL in the new fish instance.
Also tweak the SHLVL logic to interpret an environment SHLVL of "3foo"
as garbage instead of as the value "3".
Fixes#1693.
As far as I know we can't access the build artifacts from Travis, so we
can't check the interactive logs after a test failure. Add an
environment variable that causes the test runner to dump the logs
itself, and set that variable for Travis.
Split `make test` into two targets `make test_low_level` and `make
test_fishscript`, primarily so fishscript tests can be rechecked quickly
after edits.
Reformat the test.fish file and update some of the code to be a little
more straightforward (e.g. `if not cmd` instead of `if cmd; else`).
When $IFS is empty, command substitution no longer splits on newlines.
However we still want to trim off a single trailing newline, as most
commands will emit a trailing newline and it makes it harder to work
with their output.
The span now properly points at the token that was invalid, rather than
the start of the slice.
Also fix the span for `()[1]` and `()[d]`, which were previously
reporting no source location at all.
When a variable is parsed as being empty, parse out the slice and
validate the indexes anyway, behaving for slicing purposes as if the
variable had a single empty value.
Besides providing errors when expected, this also fixes the following:
set -l foo
echo "$foo[1]"
This used to print "[1]", now it properly prints nothing.
Double expansions of variables had the following issues:
* `"$$foo"` threw an error no matter what the value of `$foo` was.
* `set -l foo ''; echo $$foo` threw an error because of the expansion of
`$foo` to `''`.
With this change, double expansion always works properly. When
double-expanding a multi-valued variable, in a double-quoted string the
first word of the inner expansion is used for the outer expansion, and
outside of a quoted string every word is used for the double-expansion
in each of the arguments.
> set -l foo bar baz
> set -l bar one two
> set -l baz three four
> echo "$$foo"
one two baz
> echo $$foo
one two three four
Enhance the `read` builtin to support creating an array with the --array
flag. With --array, only a single variable name is allowed and the
entire input is tokenized and placed into that variable as an array.
Also add custom behavior if IFS is empty or unset. In that event, split
the input on every character, instead of the previous behavior of doing
no splitting at all.
One of the tests was using `>/dev/null` to suppress the `type` output.
That needs to be `^/dev/null` now, but instead just go ahead and use the
new `-q` flag.
