The block stack is now sound, and no longer needs this ancient
cleanup logic, which tried to account for cases where blocks
were pushed but never popped.
Currently the block stack is just a vector of pointers.
Clients must manually use new() to allocate a block, and then
transfer ownership to the stack (so must NOT delete it).
Give the parser itself responsibility for allocating blocks too,
so that it takes over both allocation and deletion. Use unique_ptr
to make deletion less error-prone.
Previously we would try to walk all the blocks (from within the
signal handler!) and mark them as skipped. Stop doing that, it's
wildly unsafe.
Also rationalize how the skip flag is set per block. Remove places
that shouldn't set it (e.g. break and continue shouldn't set skip
on the loop block).
read_in_chunks does not clear the intermediate string 'str'
between iterations, so every chunk has every other chunk prepended
to it.
A secondary issue is that it calls str2wcstring() on an intermediate
chunk, which may split multi-byte sequences. This needs to be deferred
to the end.
Test added. Fixes#3756
This implements a way to use the `functions` command to perform
introspection to learn about the characteristics of a function. Such as
where it came from.
Fixes#3295
If the kernel reports a size of zero for the rows or columns (i.e., what
`stty -a` reports) fall back to the `COLUMNS` and `LINES` variables. If
the resulting values are not reasonable fallback to using 80x24.
Fixes#3740
Refactor `builtin_read()` to split the code that does the actual reading
into separate functions. This introduces the `read_in_chunks()` function
but in this change it is just a clone of `read_one_char_at_a_time()`. It
will be modified to actually read in chunks in the next change.
Partial fix for #2007
The shell was doing a log of signal blocking/unblocking that hurts
performance and can be avoided. This reduced the elapsed time for a
simple benchmark by 25%.
Partial fix for #2007
We should never use stdio functions that use stdout implicitly. Saving a
few characters isn't worth the inconsistency. Too, using the forms such
as `fwprintf()` which take an explicit stream makes it easier to find
the places we write to stdout versus stderr.
Fixes#3728
If the tty has been closed (i.e., become invalid) the `ttyname()`
function will return NULL. Passing that NULL to `strstr()` can crash
fish which means it won't kill its child processes and exit cleanly.
Another fix for #3644
mkostemp is not available on some older versions of macOS. In order
for our built binaries to run on them, mkostemp must be weak-linked.
On other systems, we use the autoconf check.
Introduce a function fish_mkstemp_cloexec which uses mkostemp if
it was detected and is available at runtime, else falls back to
mkstemp. This isolates some logic that is currently duplicated in
two places.
See #3138 for more on weak linking.
In order to use C++11 with the standard macOS Xcode toolset,
we must use libc++. This in turn requires using 10.7 as our
MIN_REQUIRED in the availability macros, which in turn marks
certain wide-character functions as strong symbols (since they
were introduced in 10.7).
Redeclare them as weak, so that we can run on 10.6 without link
errors. See #3138 for more.
GNU systems don't allow mixing narrow and wide IO, so some of these
messages were lost since 1621fa43d8.
stderr is also the more logical place for error output to end up.
Fixes#3704.
Emitting warnings about EPIPE errors when writing to stdout or stderr is
more annoying than helpful. So suppress that specific warning message.
Fixes#2516
A third-party plugin noticed that using `$CMD_DURATION` in the prompt
causes problems when combined with the recent changes to tighten up
parsing of strings meant to be integer values. This fixes the problem by
ensuring the var is defined before the first interactive command is run.
See https://github.com/fisherman/dartfish/issues/7
It was pointed out that the previous change to alert people to the fact
their completion scripts were using flags that are no longer valid
resulted in way too many warnings. This limits the warning to one per
session.
Fixes#3640
On some platforms, notably GNU libc, you cannot mix narrow and wide
stdio functions on a stream like stdout or stderr. Doing so will drop
the output of one or the other. This change makes all output to the
stderr stream consistently use the wide forms.
This change also converts some fprintf(stderr,...) calls to debug()
calls where appropriate.
Fixes#3692
* Add italics and dim modifier to set_color
* update documentation for set_color
* add reverse mode to set_color
* Use standout mode as fallback for reverse mode
* Apply patch from @Darkshadow2 adding additional modes
I noticed that universal variable tests were failing on Cygwin and
Dragonfly BSD. The failures were because we are attempting to verify the
correct behavior of mechanisms that are known to be broken on those
platforms. There are still uvar test failures on those platforms with
this change but they are due to actual problems rather than bugs in the
tests.
Fixes#3587
Using `\e` is clearer and shorter than `\x1b`. It's also consistent with how
we write related control chars; e.g., we don't write `\x0a` we write '\n'.
Update our implementation of the PROMPT_SP heuristic to match current
zsh behavior. This makes it behave better on terminals like ConEmu and
the native MS Windows console which automatically insert a newline when
writing to the last column of the line.
Fixes#789
There are several places that use writestr() which should instead be
using fwprintf() or equivalent. Also, clarify the documentation for why
writestr() and writechr() exist so they aren't used inappropriately
again.
Fixes#3657
The complete builtin had once -A / --authoritative and -u /
--unauthoritative switches which indicated whether all possibilities for
completion are specified and would cause an error if the completion was
authoritative and an unknown option was encountered.
This feature was functionally removed during one of the past parser
rewritings, but -A and -u still remained in parts of the code and
command completions, although having no effect.
This commit removes the leftovers and prints an warning whenever user
tries to run the complete command with -A / -u / --authoritative /
--unauthoritative switches.
Fixes#3640.
Commit 8d27f81a to change how background jobs are handled (killed rather
than left running) when the shell is exited did not correctly handle
the nested interactive context created by the `breakpoint` command. This
fixes that mistake. Now any background jobs that already existed, or were
created within the `breakpoint` context, are left running when exiting
that context.
Fish is not consistent with other shells like bash and zsh when exiting
an interactive shell with background jobs. While it is true that fish
explicitly claims no compatibility with POSIX 1003.1 this is an area
where deviation from the established practice adds negative value.
The reason for the current behavior seems to be due to two users who did
not understand why interactive shells managed background jobs as they
did and were not aware of tools like `nohup` or `disown`. See issue
There is also a fairly significant bug present due to a misunderstanding of
what a true value from `reader_exit_forced()` means. This change corrects
that misunderstanding.
Fixes#3497
The recent discussion around allowing the user to change various termios
(i.e., stty) settings reminded me that there are places in our code
where we assume the interrupt key is [ctrl-C]. That's a bad assumption.
Instead use the actual value reported to us by the kernel.
This also makes the fkr program friendlier by always reporting when a
signal was received, not just when run with -d2, and prompting the user
to press the INTR or EOF key a second time to exit.
If acquiring a lock on the history or uvar file takes more than 250 ms
disable locking of the file. On systems with broken remote file system
locking it can cause tens of seconds delay after running each command
which can make the shell borderline unusable.
This also changes history file locking to use flock() rather than
fcntl() to be consistent with uvar file locking. It also implements the
250 ms time limit before giving up on locking.
Fixes#685
If an interactive shell has its tty invalidated attempts to write to
stdout or stderr can trigger this bug:
https://sourceware.org/bugzilla/show_bug.cgi?id=20632
Avoid that by reopening the stdio streams on /dev/null if we're getting
an ENOTTY error when trying to do things like give or take ownership of
the tty.
This includes some unrelated style cleanups but including them seems
reasonable.
Fixes#3644
When I refactored the code to reduce redundancy and improve the error
messages when the config or data directories could not be used I botched
the customization of the $HOME based data path.
min_width dates back to the original full-screen pager.
After some careful inspection, the code path that uses min_width
is never executed and so the min_width machinery is useless.
Let's remove it!
Tests that exercise error paths may result in output to
stderr. This may make it look like the test failed when it did
not. Introduce should_suppress_stderr_for_tests() to suppress
this output so the test output looks clean.
This commit fixes a bug which causes that
fish -c ')'; echo $status
("Illegal command name" error) returns 0. This is inconsistent with
e.g. when trying to run non-existent command:
fish -c 'invalid-command'; echo $status
("Unknown command" error) which correctly returns 127.
A new status code,
STATUS_ILLEGAL_CMD = 123
is introduced - which is returned whenever the 'Illegal command name *'
message is printed.
This commit also adds a test which checks if valid commands return 0,
while commands with illegal name return status code 123.
Fixes#3606.
On Solaris, some standard wide character functions are only contained in
the std:: namespace. The configure script now checks for these, enabling
the appropriate `uses` statements in src/common.h.
The checks are handwritten, because Autoconf's AC_CHECK_FUNC macro
always uses C linkage, but the problem only appears under C++ linkage.
Work on #3340.
This change increases the amount of useful information when fish is
unable to create or use its config or data directory. We now make it
clear when neither var is set or one is set to an unusable location.
Fixes#3545
After 'x' is used to delete a character at the end of a line the cursor
should be repositioned at the last character, i.e. repeatedly pressing
'x' in normal mode should delete the entire string.
A couple things went wrong with `env -u HOME USER=x ./fish -c ''`
We failed to check that `pw` isn't NULL leading to a crash when USER is
bogus. After fixing that we were not left with both variables in a
correct state still.
We now go back and force fish to dig up a working USER when we notice
this and then get both set successfully. Fixes#3599
This augments the previous change for issue #3346 by adding an error
message when an invalid integer is seen. This change is likely to be
controversial so I'm not going to squash it into the previous change.
The `test` builtin currently has unexpected behavior with respect to
expressions such as `'' -eq 0`. That currently evaluates to true with a
return status of zero. This change addresses that oddity while also
ensuring that other unusual strings (e.g., numbers with leading and
trailing whitespace) are handled consistently.
Fixes#3346
Only in one instance would test as `[` have the the errors formatted
as "[: foo". This fixes that. When trying to track down the source of
an error this could lead someone astray.
The existing code is inconsistent, and in a couple of cases wrong, about
dealing with strings that are not valid ints. For example, there are
locations that call wcstol() and check errno without first setting errno
to zero. Normalize the code to a consistent pattern. This is mostly to
deal with inconsistencies between BSD, GNU, and other UNIXes.
This does make some syntax more liberal. For example `echo $PATH[1 .. 3]`
is now valid due to uniformly allowing leading and trailing whitespace
around numbers. Whereas prior to this change you would get a "Invalid
index value" error. Contrast this with `echo $PATH[ 1.. 3 ]` which was
valid and still is.
Builtin commands that validate var names should use a consistent
mechanism. I noticed that builtin_read() had it's own custom code that
differed slightly from wcsvarname().
Fixes#3569
My previous change removed one place where is_wchar_ucs2() was used and
replaced it with compile time tests. This change does the same for the
other uses.
On Cygwin there are two narrowing conversions at line 931 in
src/fish_tests.cpp due to the code assuming a wchar_t is four bytes.
Obviously that's wrong but only became an issue with the pending change to
switch to C++11. The problematic values aren't actually used on Windows
because the tests that would use them are bypassed if is_wchar_ucs2()
returns true. This change predicates that code on a compile time rather
than a run time test.
There isn't a good reason to disallow an explicitly empty completion
description. Since I'm touching the code also modify the argument
parsing the match the style of most of the builtins.
Fixes#3557.
This fixes some of the IWYU and cppcheck lint warnings. And only on
macOS (formerly OS X). Fixing these types of warnings on a broader set
of platforms should be done but this is a baby step to making `make
lint-all` have few, if any, warnings. This reduces the number of lines
in the `make lint-all` output on macOS by over 500 lines.
Switch from a linear to a binary search when looking for a matching
string in an enum map. Testing shows this is a little more than twice as
fast when searching for keywords in the sixteen entry keyword_map array.
This speedup doesn't matter much when searching for subbcommands but any
slow down in the parser is unacceptable.
I'm going to use the same mechanism elsewhere such as token_type_map
in src/parse_tree.cpp. But this change only affects the recently
introduce subcommand handling for the history and status commands.
Verified on Cygwin on MS Windows 7 when invoked as
`env LANG=zh_CN.GBK@cjknarrow fish`. No regression seen
when run on other systems with UTF-8 locales.
Fixes#3503
The `status` command currently silently allows incompatible flags (i.e.,
subcommands). Too, using flags to specify subcommands misleads the user
into thinking they can specify multiple subcommands.
We recently modified the `history` command to deprecate using flags for
subcommands. This change does the same for the `status` command.
Fixes#3509
Earlier lint cleanups overlooked a couple of modules because on macOS at
the moment oclint ignores them. I noticed this when I ran `make lint-all`
on Ubuntu.
My earlier attempt with commit 851e449 to eliminate all the compiler
warnings about mixing signed and unsigned ints in an expression
introduced a subtle bug. This fixes that mistake.
Fixes#3488
There was one block of code modified by commit 42458ff7 that had
convoluted, inverted, logic. In the process of collapsing nested
"if" blocks the logic was modified to avoid using "!" everywhere the
bool was tested. Unfortunately I neglected to modify two of the
conditions used to set that var to reflect the changed polarity.
The fish_key_reader program was the only user of the
`set_wait_on_escape_ms()` function and that use was removed with commit
0461743. So remove it from the main fish code. This was found by `make
lint`.
Update the CHANGELOG to more accurately reflect what will be included in
the 2.4.0 release vis-a-vis the `history` command behavior.
I noticed that the compiler was emitting some harmless warnings related
to the history changes so deal with those as well.
This modifies the code path for `set PATH` and `set CDPATH` to emit an
easier to understand warning when an entry in those vars is invalid. For
example
$ set PATH $PATH /tmp/arglebargle
set: Warning: $PATH entry "/tmp/arglebargle": No such file or directory
$ mkdir /tmp/d
$ chmod 0 /tmp/d
$ set PATH $PATH /tmp/d
set: Warning: $PATH entry "/tmp/d": Permission denied
$ touch /tmp/x
$ set PATH $PATH /tmp/x
set: Warning: $PATH entry "/tmp/x": Not a directory
Fixes#3450
* Fix building on Android by avoiding getpwent() if missing with autoconf check
The getpwent() function does not link when building for Android,
and user names on that platform are not interesting anyway.
Using a configure check for stat.st_ctime_nsec fixes building on
Android which has that field but does not define STAT_HAVE_NSEC.
Before this change the Android build failed on the st_ctim.tv_nsec
fallback #else clause.
Taking a different approach here. I can't see why we'd only want to
recognize certain colors. Now, we'll just try all the colors fish might
use.
This could probably be optimized now that there are more
than 8 (or 16) colors fish can do.
It is believed there are no longer any platforms we support that do not
support passing NULL as the second argument to realpath(). So rather
than duplicating the logic to get reasonable behavior from our
wrealpath() wrapper simply remove the redundant implementation.
After implementing `builtin fish_realpath` it was noticed that it did
not behave like GNU `realpath` without options. Which is super annoying
since that was the whole point of implementing the command. Major
failure on my part since I wrote the unit tests to match the behavior of
the existing `wrealpath()` function that I simply exposed as a builtin
command. Rather than actually verifying it behaved in a manner
compatible with GNU realpath.
Also, while the decision to call the builtin `fish_realpath` seemed to
make sense at the time of the original commit further reflection has
shown that to be a silly, idiosyncratic, thing to have done. So rename
it to simply `realpath`.
Fixes 3400
It's the ninth color - on virtual consoles this was likely to
try a color that doesn't work because we checked if max_colors >= 8.
Add another way to reach that color on terminals with only 8 colors
by using bold mode to get a bright.
This has potential to fail by simply rendering as black which can cause
it to be invisible on a white-on-black terminal. Not bad as it's just
making this bell/whistle invisible:
We *really* want to set the omitted newline character apart by having
it appear grey. On (FreeBSD consoles, at least) VCs it's not uncommon
for it to render as a "?". It's particularly confusing if it doesn't
render in a darker color as it cannot be discerned from actual program
output.
When performing fuzzy completion, if a directory segment is
valid, then don't consider it for a fuzzy match even if
the literal match produces no results.
Fixes#3211
The use of wcstoimax causes certain out-of-range values
to be silently truncated (e.g. when converted to a pid),
and is incompatible with FreeBSD (see #626)
This reverts commit 6faa2f9866.
The template has different behavior around interpreting
non-decimal sequences. This doesn't seem to have been intended.
This reverts commit f843eb3d31.
Both GNU and BSD have bugs regarding the classification of
non-characters and private use area characters. Provide wrappers around
iswalnum(), iswalpha(), and isgraph() to provide a consistent
experience. We don't bother to autoconf the use of these wrappers for
several reasons. Including the fact that a binary built for one distro
release should behave correctly on another release (e.g., FreeBSD 10
does the right thing while FreeBSD 11 and 12 do not with respect to
iswalnum() of code points in the range 0xFDD0..0xFDFF).
Also move a few functions from common.* to wutil.* because they are wide
char specific and really belong in the latter module.
Fixes#3050
