If the read thread has already exited, it will have closed the read end
of the quit pipe. Unless SIGPIPE is masked with signal(SIGPIPE, SIG_IGN),
or the macOS-specific fcntl(F_SETNOSIGPIPE), writing to the write end of
a broken pipe kills the writer with SIGPIPE instead of returning -EPIPE
as an error. This causes a crash if the read thread exits before
threadExit.
This was already a possible race condition if read() returns
error.NotOpenForReading or error.InputOutput, but it's now much easier to
trigger due to the recent "termio/exec: fix 100% CPU usage after
wait-after-command process exits" fix.
Fix this by closing the quit pipe instead of writing to it.
'--posix' starts bash in POSIX mode (like /bin/sh). This is rarely used
for interactive shells, and removing automatic shell integration support
for this option allows us to simply/remove some exceptional code paths.
Users are still able to manually source the shell integration script.
Also fix an issue where we would still inject GHOSTTY_BASH_RCFILE if we
aborted the automatic shell integration path _after_ seeing an --rcfile
or --init-file argument.
The renderer must track if the foreground, background, and cursor colors
are explicitly set by an OSC so that changes are not overridden when the
config file is reloaded.
`std.fs.accessAbsolute` asserts if the user proposed path is absolute,
which we are seemingly passing as-is with no validating that it is.
When running with safety checks on, passing non-absolute path to
--working-directory will make ghostty crash.
I changed it to use `Dir.access`, which is just `accessAbsolute` without
the check.
This has the side effect of also allowing relative working directory.
We're packaging more and more application-specific data directories in
our application bundle. It's helpful to add that path to XDG_DATA_DIRS
so those applications (that support XDG_DATA_DIRS) can locate their data
directories without additional user-level configuration.
This also fixes a typo ("MATHPATH") in the nearby MANPATH-building code.
Bash shell detection was originally disabled in #1823 due to problems
with /bin/bash on macOS.
Apple distributes their own patched version of Bash 3.2 on macOS that
disables the POSIX-style $ENV-based startup path:
e5397a7e74/bash-3.2/shell.c (L1112-L1114)
This means we're unable to perform our automatic shell integration
sequence in this specific environment. Standard Bash 3.2 works fine.
Knowing this, we can re-enable bash shell detection by default unless
we're running "/bin/bash" on Darwin. We can safely assume that's the
unsupported Bash executable because /bin is non-writable on modern macOS
installations due to System Integrity Protection.
macOS users can either manually source our shell integration script
(which otherwise works fine with Apple's Bash) or install a standard
version of Bash from Homebrew or elsewhere.
Fixes#2857
Some terminal modes always reset, but there are others that should be
conditional based on how the terminal's default state is configured.
Primarily from #2857 is the grapheme clustering mode (mode 2027) which
was always resetting to false but should be conditional based on the
the `grapheme-width-method` configuration.
Make the foreground_color and background_color fields in the Terminal
struct optional values so that we can determine if a foreground or
background color was explicitly set with an OSC 10 or OSC 11 sequence.
This makes the logic a bit simpler to reason about (i.e.
`foreground_color` is now always "the color set by an OSC 10 sequence"
while `default_foreground_color` is always "the color set by the config
file") and also fixes an issue where an OSC 10 or OSC 11 query would not
report the correct color after a config update changed the foreground or
background color.
The `cursor_color` field was already optional, with the same semantics
(it is only non-null when explicitly set with an OSC 12) so this brings
all three of these fields into alignment.
Related #2755
From the mode 2031 spec[1]:
> Send CSI ? 2031 h to the terminal to enable unsolicited DSR (device status
> report) messages for color palette updates and CSI ? 2031 l respectively to
> disable it again.
>
> The sent out DSR looks equivalent to the already above mentioned. This
> notification is not just sent when dark/light mode has been changed by the
> operating system / desktop, but also if the user explicitly changed color
> scheme, e.g. by configuration.
My reading of this paired with the original discussion is that this is
meant to be sent out for anything that could possibly change terminal
colors.
Previous to this commit, we only sent out the DSR when the actual system
light/dark mode changed. This commit changes it to send out the DSR on
any operation that _may_ change the terminal colors.
[1]: https://contour-terminal.org/vt-extensions/color-palette-update-notifications/#example-source-code
When an empty string is given to OSC7, the pwd is reset to nil (as if
the terminal never received a pwd report to begin with). This is
analogous to how OSC0/2 reset the title to nil when given an empty
string.
This is practically useful for macOS because it allows our proxy icon to
also be reset instead of being stuck on the last known path.
This breaks from any known terminal behavior. As far as I can find, this
is totally unspecified so we're somewhat free to do what we want. I
don't think any terminal programs depend on this behavior, so I think
it's safe to change it.
Fixes#2651
First, our OSC parser didn't allow blank OSC 0 or 2 requests. This
should be allowed and this fixes that with a test.
Second, it seems many terminals (iTerm2, Kitty) treat setting a blank
title as resetting to whatever the default title is rather than
explicitly setting it as blank. If a program wants a title to be blank
they should send a single space. This commit follows this behavior.
The explanation now refers to RFC 793 instead of just claiming some
arbitrary value as truth. The previous value was correct, but now there
is a proper source for the correct value.
Mode 2048 and CSI 14 t are size report control sequences which contain
the text area size in pixels. The text area is defined to be the extents
of the grid (rows and columns). Ghostty calculates the available size
for the text area by setting the available padding, and then filling as
much of the remaining space as possible. However, if there are remainder
pixels these are still reported as part of the text area size.
Pass the cell_size geometry through so that we can always report the
correct value: columns * cell width and rows * cell height.
Fixes#2273
On macOS, killpg is expected to fail with EPERM because of the way we
launch a login process around it. Before this commit, this caused us to
never call waitpid and reap the child process, which caused the child
process to stick around as a zombie.
This commit allows killpg to fail with EPERM on macOS and fall through
to waitpid.
