Remove isatty() check for stdout - this was added for both stdout
and stdin because "there is no reason to do that", but there is one:
Leaves only the bind command printed ot stdout, this allows
for one to do `fish_key_reader > bind_command.fish` to capture the bind
command while seeing the rest of the output.
After the colorized syntax output in type -a foo, "foo is /usr/..."
would also be colored. (or 'test' in fish_indent foo.fish; echo test).
Make fish_indent reset the color when it's done.
I did some research and experiments. For good or bad the `bind` command
requires the use of wide char codepoints (e.g., \u1234) for non-ASCII
chars. So don't force the use of the POSIX locale, but do provide it as
an option for people who want to see the individual bytes rather than a
decoded wide char.
Simplify the format of the information displayed for each character. There
really isn't much point in providing decimal, octal, and hexadecimal. Just
print hex and symbolic representations.
Add an example `bind` command that a user can copy/paste.
Closes#3183
Another developer noticed that redirecting stdin of `fish_key_reader`
results in weird behavior. Which is not at all surprising. So add checks
to ensure stdin and stdout are attached to a tty.
Add some rudimentary unit tests for this program.
A discussion on Gitter proposed allowing the user to signal their desire to
exit fish_key_reader by pressing \cC or \cD twice in a row. This implements
that.
I also decided to refactor how signals are handled. Most notably receiving a
signal will no longer print a diagnostic message unless you've enabled
debugging with `-d2` (or higher level).
In addition to fixing the setting of the locale to C/POSIX this also
corrects several problems introduced by the commits made in the past
couple of days. As a consequence of dealing with all of this I decided
to refactor the code to simplify one of the overly long functions I
introduced in my previous change.
Fixes#3168
There is no conceivable way in which timef()'s invocation of gettimeofday()
can fail where it makes sense to continue running. Yes, one such,
legitimate, failure mode is a 32-bit kernel and the date is greater than
2038-01-19 03:14:07. If you're running a fish binary on such a system
it's time to upgrade. Otherwise, either the hardware or OS is broken.
Fixes#3167.
* Correct notice about ^C
* Move time deltas to end of the line away from the important info on
left.
* Use timef() instead of gettimteofday() ourselves
* Show time in ms (is this even useful in any unit? Maybe testing escape
delays...)
* Make init more similar to other apps.
```
~ $ set -e TERM; fish
Assertion failed: (!is_missing), function c_str, file src/env.cpp, line 690.
fish: 'fish' terminated by signal SIGABRT (Abort)
```
The tty device timestamps on MS Windows aren't usable because they're always
the current time. So fish can't use them to decide if the entire prompt needs
to be repainted.
Fixes#2859
* if (result == ULLONG_MAX) is always false, likely a typo as
result is unsigned long, and the comment says ULONG_MAX.
* use off_t instead of size_t for file size where it can mismatch
st_size's type in stat.h
For example, an argument 12345^ is a real argument, not a redirection
There's no reason to use ^ here instead of >, and it's annoying to git
users.
Fixes#1873
When given no path, the logic was happy to try to use
an unitialized output_location.
$ fish_indent -w < test.fish
Opening "(null)" failed: Bad address
Initialize the string, and repair the logic to catch this case
and report the problem correctly.
Update Xcode project, HeaderDoc comments.
Fix various invalid HeaderDoc comments. Normalize autoload.cpp/autoload.h as an example of something closer to "proper" HeaderDoc formatting.
Have clang/Xcode validate HeaderDoc comments. Remove key_reader.cpp from Xcode project.
Fix test setup bogosities. Specifically, they weren't hermetic with respect to
locale env vars.
Rewrite the handling of locale vars to simplify the code and make it more like
the pattern most programs employ.
Fixes#3110
We need to actually export the curses/terminfo env vars in order for
`setupterm()` to be able to use them. While fixing this I reworked the
fallback logic implemented by @zanchey in response to issue #1060 in
order to simplify the logic and clarify the error messages.
This does not allow someone to change the curses/terminfo env vars after
the first prompt is displayed (you can but it won't affect the current
fish process). It only makes it possible to set `TERM`, `TERMINFO`, and
`TERMINFO_DIRS` in *config.fish* or similar config file and have them be
honored by fish.
The issue here is that when inserting a common prefix for e.g. a
substring match, we increase the amount of available candidates again to
things the user didn't want.
An example is in share/functions - a completion for "inter" would
previously expand to "__fish_" because it matched:
- __fish_config_interactive.fish
- __fish_print_interfaces.fish
- __fish_print_lpr_printers.fish
The completion afterwards would then show 189 possible matches, only
three of which (the above) actually matched the original "inter".
Fixes#3089.
Cppcheck was complaining about the `return val.c_str()` at the end of the
`wgettext()` function. That would normally a bug since the lifetime of
`val` ends when the function returns. In this particular case that's not
true because the string is interned in a cache. Nonetheless, rather than
suppress the lint warning I decided to modify the API to be more idiomatic.
In the process of fixing the aforementioned lint warning I fixed several other
lint errors in that module.
This required making our copy of `wgetopt()` compatible with the rest of
the fish code. Specifically, by removing its local definitions of the
"_" macro so it uses the same macro used everywhere else in the fish
code. The sooner we kill the use of wide chars the better.
This only eliminates errors reported by `make lint`. It shouldn't cause any
functional changes.
This change does remove several functions that are unused. It also removes the
`desc_arr` variable which is both unused and out of date with reality.
This makes the wide char tests run by `./fish_tests` pass on systems where
sizeof wchar_t is two (e.g., Cygwin). In doing so it corrects several
problems with the underlying code in module *utf8.cpp* such as allowing
five and six byte UTF-8 sequences. They were allowed by the original
Unicode proposal but are not allowed by the adopted standard.
Configure the tty driver to ignore the lnext (\cV) and werase (\cW) characters
so they can be bound to fish functions.
Correct the `fish_key_bindings` program to initialize the tty in the same
manner as the `fish` program.
Fixes#3064
Overwriting the user's clipboard by default is annoying and contributors
don't use it.
This is better served via an explicit binding that calls e.g. `xsel`.
This change allows the user to specify the script name on the CLI in addition
to being redirected from stdin. It also adds a `-w` flag to write the modified
script to the original file.
I noticed that the `test_convert()` function was randomly failing when
run on OS X Snow Leopard. I tracked it down to the `mbrtowc()` function on
that OS being broken. Explicitly testing for UTF-8 prefixes that identify
a sequence longer than four bytes (which the Unicode standard made illegal
long ago) keeps us from having encoding errors on those OS's.
This also makes the errors reported by the `test_convert()` function actually
useful and readable.
Lastly, it makes it possible to build fish on OS X Snow Leopard.
Drops configure check for wcsdup, wcslen, wcscasecmp, wcsncasecmp,
wcwidth, wcswidth, wcstok, fputwc, fgetwc, and wcstol. Drop the fallback
implementations of these on non-Snow Leopard platforms.
Work on #2999.
fwprintf would segfault on DragonFly BSD 1.4.0, released in January
2006. This was fixed by DragonFly BSD 1.4.4, released in April 2006. It
seems unlikely that anyone is still running a ten-year-old, unsupported
version, and hoping that fish will continue to build.
I've checked this in virtual machines.
Work on #2999.
The autoconf-generated config.h contains a number of directives which
may alter the behaviour of system headers on certain platforms. Always
include it in every C++ file as the first include.
Closes#2993.
This change does several things. First, and most important, it allows
dumping the "n" most recent stack frames on each debug() call. Second,
it demangles the C++ symbols. Third, it prepends each debug() message
with the debug level.
Unrelated to the above I've replaced all `assert(!is_forked_child());`
statements with `ASSERT_IS_NOT_FORKED_CHILD()` for consistency.
Fix a minor bogosity I noticed while building fish on OS X Snow
Leopard. It's technically not a bug because only old compilers complain
about the original statement but this change makes the one line this
changes consistent with the rest of the fish code.
It's currently too easy for someone to bork their shell by doing something
like `function test; return 0; end`. That's obviously a silly, contrived,
example but the point is that novice users who learn about functions are
prone to do something like that without realizing it will bork the shell. Even
expert users who know about the `test` builtin might forget that, say, `pwd`
is a builtin.
This change adds a `--shadow-builtin` flag that must be specified to
indicate you know what you're doing.
Fixes#3000
I'm doing this as part of fixing issue #2980. The code for managing tty modes
and job control is a horrible mess. This is a very tiny step towards improving
the situation.
The original `key_reader` program was useful but didn't do much that `xxd`
or `od -tx1z` didn't do. Furthermore, it wasn't built and installed by
default. This change adds features that make it superior to those programs
for decoding interactive key presses and makes it a first-class citizen
like the `fish_indent` program that is always available.
Fixes#2991
The fork (create new process) related debugging messages rely on an
undocumented env var and use `printf()` rather than `debug()`. There are
also errors in how the fork count is tracked that this fixes.
Fixes#2995
Some `oclint` errors regarding "useless parentheses" are meaningfull. But
the vast majority are bogus in as much as removing the parentheses reduces
readability. So fix a few of the egregious uses and otherwise suppress
that error.
I missed restyling a few "switch" blocks to make them consistent with the rest
of the code base. This fixes that oversight. This should be the final step in
restyling the C++ code to have a consistent style. This also includes a few
trivial cleanups elsewhere.
I also missed restyling the "complete" module when working my way from a to z
so this final change includes restyling that module.
Total lint errors decreased 36%. Cppcheck errors went from 47 to 24. Oclint P2
errors went from 819 to 778. Oclint P3 errors went from 3252 to 1842.
Resolves#2902.
For this change I decided to bundle the remaining modules that need to be
resytyled because only two were large enough to warrant doing on their own.
Reduces lint errors from 225 to 162 (-28%). Line count from 3073 to 2465 (-20%).
Another step in resolving issue #2902.
Reduces lint errors from 36 to 33 (-8%). Line count from 1910 to 1476 (-23%).
Another step in resolving issue #2902.
This also fixes a stupid mistake from an earlier commit where I didn't realize
that osx/config.h was meant to be included as a semi-static file in the
repository.
fish_title currently outputs some escaped text, which can confuse
the line driver (#2453). Issue a carriage return so the line driver
knows we are at the beginning of the line, unless we are writing
the title as part of the prompt. In that case, we may have text from
the previous command still on the line and we don't want to move the
cursor.
Fixes#2453
Don't `#include "*.cpp"` modules in other cpp modules. I already took care
of all the builtin_*.cpp modules in my previous change where I restyled
the builtin code. This change fixes the two remaining instances of this
anti-pattern.
Now that the IWYU cleanup has been merged compile all, not just a couple, of
the builtin modules independent of builtin.cpp. That is, no longer `#include
builtin_NAME.cpp` in builtin.cpp. This is more consistent, more in line with
what developers expect, and is likely to reduce mistakes.
Reduces lint errors from 384 to 336 (-13%). Line count from 6307 to 4988 (-21%).
Another step in resolving issue #2902.
- Set PCRE2_SUBSTITUTE_OVERFLOW_LENGTH to get the required buffer length
from pcre2 instead of guessing
- Set PCRE2_SUBSTITUTE_EXTENDED to enable extra goodies in the
replacement string
Remove the "make iwyu" build target. Move the functionality into the
recently introduced lint.fish script. Fix a lot, but not all, of the
include-what-you-use errors. Specifically, it fixes all of the IWYU errors
on my OS X server but only removes some of them on my Ubuntu 14.04 server.
Fixes#2957
The OS X Xcode IDE has a weird requirement that block comments preceding a
function or class definition must begin with three slashes rather than two if
you want the comment displayed in the "Quick Help" window.
Make the history code conform to the new style guide. Every change was
produced by clang-format (e.g., `make style`) with the exception of
comments which were manually reformatted. That has to be done by hand
since clang-format leaves comments alone other than to reflow comment
lines to get them below the allowed line length.
The total number of lines is reduced by 313 lines (13%) in the two
affected files. Line count is generally a poor metric but in this
case it reflects an increase in information density without a loss in
readability. Furthermore, the standardization of braces, whitespace,
and comment style will make it easier for people to read the code.
This reduces the number of warnings by `make lint` from 168 to 87 (a 48%
decrease). Making it much easier to focus on the substantive lint issues.
Further improvements are possible. For example, many comments are not
very helpful (e.g., they point out the obvious) or provide insufficient
detail. But those are beyond the scope of this change.
This is the first step in resolving issue #2902.
The readlink() function does not null terminate the path it returns.
Remove the OS X code that deals with a path buffer that is too short. For
one thing a loop isn't needed since we're told how big of a buffer
is required if the first _NSGetExecutablePath() call fails. But more
important it is so unlikely that the path will be longer than PATH_MAX
that if it is we should just give up.
Fixes 2931.
I just noticed that depending on the state of your working tree there can be
one or more spaces after the modification token and the file name. If there is
more than one space that causes the `string split` to produce unexpected
output.
In keeping with the change made by @ridiculousfish earlier today modify
the `keyword_description()` function to return a const wchar_t pointer.
Also, simplify the `token_type_description()` function to use the recently
introduced mapping array. This changes the wording of many of the token
type descriptions. However, I can't see this as being a problem since
the original descriptions (e.g., "token_redirection") are no clearer to
someone not acquainted with the implementation.
Fish keywords can be quoted and split across lines. Prior to this change
`fish_indent` would retain such odd, obfuscated, formatting. This change
results in all keywords being converted to their canonical form.
This required fixing a bug: the keyword member of parse_node_t wasn't being
populated. This hadn't been noticed prior to now because it wasn't used.
Fixes#2921
This code represents only risk and does nothing useful for anything
that can compile fish.
In C++ situations where __STDC_VERSION__ is unset (as it should be),
fish was assuming we are on < C99 and setting it to __FUNCTION__.
Basically always, __PRETTY_FUNCTION__ ends up reaplaced by __FUNCTION__, this hurt
error message usefulness and richness.
__PRETTY_FUNCTION__: const thing::sub(int)
__FUNCTION__: sub
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
When determining the old path, get the existing value in any scope,
not just the set scope. Also only complain about absolute paths:
relative paths are expected to be invalid sometimes.
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
Per discussion in pull-request #2891, it's not available on Linux (we just
fill it with zero), and unless run as root on OS X (or other BSD system) it
will be zero. Remove it from file_id_t. Also fix the initialization of the
file_id_t structure.
Fixes#2891
This is a quick and dirty conversion of the atypical, and undocumented,
logging done by env_universal_common.cpp to the usual `debug()` pattern. I
didn't want to drop the messages because they could be useful when
debugging future issues. So I simply converted them to the lowest debug
level using the normal debug() function.
Fixes#2887
Cppcheck has identified a lot of unused functions. This removes funcs that
are unlikely to ever be used. Others that might be useful for debugging I've
commented out with "#if 0".
This fixes all memory leaks found by compiling with
clang++ -g -fsanitize=address and running the tests.
Method:
Ensure that memory is freed by the destructor of its respective container,
either by storing objects directly instead of by pointer, or implementing
the required destructor.
When explicitly asking for the fish version string the information
should go to stdout rather than stderr. Also, there is no reason to use
exit_without_destructors() rather than exit() in that code path. We
actually want the side-effects of exit() such as flushing stdout and
there aren't any threads or other things that could cause a normal exit
to fail when that function is run.
The early return skipped all cleanup.
This problem is a case for the classic "goto fail" paradigm, but this
change instead makes a few adjustments to take advantage of a previously
unused level of indentation to conditionally execute the success path.
The error message now prints the filename instead of "open",
which should be more idiomatic.
Tip:
This patch makes sense if viewed with `git show --ignore-space-change`.
The swap-selection-start-stop function goes to the other end of the highlighted text, the equivalent of `o' for vim visual mode.
Add binding to the swap-selection-start-stop function, `o' when in visual
mode.
Document swap-selection-start-stop, begin-selection, end-selection, kill-selection.
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.
The u_int typedef fails to compile on all platforms (e.g. Windows). It
is part of the code imported from tmux.
Update it to the SUS-standard uid_t.
Closes#2821.
Address the feedback from the prior commit:
- Change the sense of return value testing to match more common
comparison idiom
- Test result of fchmod as well as fchown
- Change sense of return value testing around wrename as well
- Include errno where possible in error message
The function fchown is annotated with warn_unused_result. As
formerly used in the code, it would emit a compiler warning
```warning: ignoring return value of ‘fchown’, declared with
attribute warn_unused_result [-Wunused-result]```
This commit notes the return value and emits appropriate error/logging
messages if the call fails, creating more traceable results and
satisfying the compiler.
There is no longer a good reason to detect whether or not getopt_long()
is available. All UNIX implementations we're likely to run on have it. And
if we ever find one that doesn't the right thing to do is not fallback to
getopt() but to include the getopt_long() source in our package like we
do with the pcre2 library. Since it's licensed under LGPL we can legally
do so if it becomes necessary.
This partially addresses issue #2790.
Previously, when decoding UTF-8, we would first run through the
array to compute the correct size, then allocate a buffer of that size,
then run through the array again to fill the buffer, and then copy it
into a std::wstring. With this fix we can copy it into the string
directly, reducing allocations and only requiring a single pass.
This narrows the range of Unicode codepoints fish reserves for its own
use from U+E000 thru U+F8FE (6399 codepoints) to U+F600 thru U+F73F (320
codepoints). This is still not ideal since fish shouldn't be using any
Unicode private-use codepoints but it's a step in the right direction.
This partially addresses issue #2684.
This was used to cache a narrow string representation
of commands, so that if certain system calls returned errors
after fork, we could output error messages without allocating
memory. But in practice these errors are very uncommon, as are
commands that have wide characters. It is simpler to do a best-effort
output of the wide string, instead of caching a narrow string
unconditionally.
Prior to this fix, read_ni would use parse_util_detect_errors
to lint the script to run, and then parser_t::eval() to execute it.
Both functions would parse the script into a parse tree. This allows
us to re-use the parse tree, improving perfomance.
Introduces a new template moved_ref which is like an rvalue reference.
This allows passing around objects while being explicit that the
receiver may acquire ownership. This will help reduce some allocations.
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
