On my system at least (fedora 15), git-symbolic-ref is an invalid
command. Not sure if it's a BIC change from git itself, a distribution
thing, or a mistake on my end. Either way, no harm in using the
extended version. Now I get git branch status (yay).
The first defaults.list file found should not override all the other ones,
it just needs to be searched first.
E.g. For me (cat ~/.local/share/applications/defaults.list) returns
[Default Applications]
text/html=chromium-browser.desktop
So this should be used for text/html mimetypes, but the other defaults.list
files should be searched for other mimetypes.
I had to refactor get_filename so that it can return all the filenames, so
I changed it to append_filenames that appends all the filenames to a list
and provided a wrapper function called get_filename.
On Gentoo the eix program is MUCH faster than emerge for listing package
names. I've left the emerge code in as a 2nd choice because not every
Gentoo system has eix installed (although they should). Also, the
emerge code didn't seem to produce any output on my system.
Signed-off-by: Grissiom <chaos.proton@gmail.com>
get_filename tried to work around this with hardcoded strings that end with
a '/', but would fail to work properly for environment variables
XDG_DATA_HOME or XDG_DATA_DIRS that don't do the same.
Renamed references to the previous command name of save_function to
funcsave, and returned an error after printing the help text when no
arguments are specified.
It is documented this way, but the common builtin_run function was
checking for -h or --help before even calling the builtin_count. Without
this functions like funced can't use count to check the number of
arguments before checking for -h or --help.
Signed-off-by: Grissiom <chaos.proton@gmail.com>
Treat fish_greeting as a whole when show up the greeting messages. And
the user may want to set fish_greeting to an null value or an array.
This requires that the variable be quoted when used as an argument for
switch in __fish_config_interactive.
Signed-off-by: James Vega <jamessan@debian.org>
[modified the commit message]
Signed-off-by: Grissiom <chaos.proton@gmail.com>
The builtin function returns the raw status, which is not the same as
the integer return by waitpid. Therefore, the WIF macros shouldn't be
used for checking or obtaining the status.
This bug can be seen by executing (eval false). The builtin eval
function returns 1, but proc_format_status misinterprets it as a signal
by checking WIFSIGNALED, so adds 128 to the status to return 129.
Signed-off-by: Grissiom <chaos.proton@gmail.com>
Fix spelling mistakes in various bits of the documentation.
Signed-off-by: James Vega <jamessan@debian.org>
Signed-off-by: Grissiom <chaos.proton@gmail.com>
Previously the expression only searched for anchors with a name
attribute, but doxygen 1.7.1 is producing anchors with the id attribute
instead. The sed expression allows both cases for compatibility.
I also used single quotes for the sed expression to avoid escaping
double quotes.
Signed-off-by: Grissiom <chaos.proton@gmail.com>
This also caused (isatty < /dev/null) to return 0 since it uses eval,
and (ls | cat) to output using the classify indicator style since it
uses isatty. This is how I found the bug.
Reviewed-by: Grissiom <chaos.proton@gmail.com>
docdir was previously being set to "${prefix}/share/doc/${PACKAGE_NAME}"
when it wasn't explicitely set on configuration using --docdir. Without
this appearing in the Makefile, some files silently get silently
installed directly into ${prefix}/share/doc instead within a fish
subdirectory.
I also added datarootdir to fix an autoconf warning, since autoconf
normally would use it for the directory paths (e.g. docdir =
${datarootdir}/doc/${PACKAGE_TARNAME}). The autoconf generated configure
script has a hack to fix this, but states: "FIXME: This hack should be
removed a few years after 2.60."
a command to accept the "copy". So, you add your clipboard buffer command
to an environment variable, and custom commands will be used for the copy
program. Very useful when your OS is not naitively supported by fish.
Way to reproduce:
> set -e TERM
fish: function contains_internal called with null value for argument a.
fish: This is a bug. If you can reproduce it, please send a bug report
to fish-users@lists.sf.net.
fish: Backtrace:
/home/grissiom/sysroot/bin/fish(show_stackframe+0x1a) [0x43519a]
/home/grissiom/sysroot/bin/fish(contains_internal+0xce) [0x4373ee]
/home/grissiom/sysroot/bin/fish(reader_write_title+0x59) [0x424179]
/home/grissiom/sysroot/bin/fish() [0x4242d5]
/home/grissiom/sysroot/bin/fish(reader_readline+0x54) [0x425974]
/home/grissiom/sysroot/bin/fish(reader_read+0xba) [0x42742a]
/home/grissiom/sysroot/bin/fish(main+0x4be) [0x43b19e]
/lib64/libc.so.6(__libc_start_main+0xfd) [0x7fcbe0dcbb6d]
/home/grissiom/sysroot/bin/fish() [0x408c29]
This commit makes fish_parse_opt stop at first non-option argument,
which expected as fish script name and pass the remaining argument to
that script. I also updated the test cases.
