This commit adds support for inserting in a "verbatim" mode where a
single uninterpreted key is appended to the buffer.
As this allows the user to input control characters, all control
characters except \n (^M) are rendered in their caret form, with
reverse video (SGR 7) applied to it.
To not break cursor movement, the concept of "masked" characters is
introduced to the StringMetrics interface, which can be mostly ignored
by the rest of the system.
It should be noted that unlike some other line editing libraries,
LibLine does _not_ render a hard tab as a tab, but rather as '^I',
which greatly simplifies cursor handling.
Other shells also support a number of other options with exec and
some have special behaviour when calling exec with no arguments except
redirections.
This PR only supports the basic case of replacing the Shell process
(or LibShell host process) with the provided command.
Such errors are raised when SyntaxError nodes are executed, and are also
used for internal control flow.
The 'break' and 'continue' commands are currently only allowed inside
for loops, and outside function bodies.
This also adds a 'loop' keyword for infinite loops.
Fixes#4345.
This was done in 54b453b in the name of "fixing event loop processing in
subshells", but I do not see how a new PGID is supposed to affect the event
loop.
This seems to have been done by mistake, let's see if any tests fail
because of this.
As Vector<T> will relocate objects to resize, we cannot assume that the
address of a specific LocalFrame will stay constant, or that the
reference will not be dangling to begin with.
Fixes an assertion that fires when a frame push causes the Vector to
reallocate.
waitid() *may* be called many times if a job does not exit, so only
assert this fact when the job has in fact exited.
Also allows Background nodes to contain non-execute nodes.
This allows us to easily re-use history loading and saving in other
programs using Line::Editor, as well as implementing universally
recognized HISTCONTROL.
In the future all (normal) output should be written by any of the
following functions:
out (currently called new_out)
outln
dbg (currently called new_dbg)
dbgln
warn (currently called new_warn)
warnln
However, there are still a ton of uses of the old out/warn/dbg in the
code base so the new functions are called new_out/new_warn/new_dbg. I am
going to rename them as soon as all the other usages are gone (this
might take a while.)
I also added raw_out/raw_dbg/raw_warn which don't do any escaping,
this should be useful if no formatting is required and if the input
contains tons of curly braces. (I am not entirely sure if this function
will stay, but I am adding it for now.)
Theoretically, this assertion should never trip (at least, on serenity
where we don't really reuse PIDs).
This change should be considered temporary, until we figure out whether
waitpid() is being called twice (and succeeding) for one given PID.
This really just works around the core issue, which is that we have no
reliable way to know exactly who raised the signal (yet).
Fixes#3645, in a very weird (yet apparently standard) way.
The following example should illustrate one issue arising from this:
$ echo 'exit 1' > example.sh
$ Shell example.sh
Good-bye!
This message is meant to be shown to an interactive user, but not in a
shell script.
