scsh-conservatory
/
scsh-0.5
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
scsh-0.5/doc/scsh-manual/running.tex

899 lines
36 KiB
TeX
Raw Blame History

%&latex -*- latex -*-
\chapter{Running scsh}
\label{chapt:running}
Scsh is currently implemented on top of {\scm}, a freely-available
{\Scheme} implementation written by Jonathan Rees and Richard Kelsey.
{\scm} uses a byte-code interpreter for good code density, portability
and medium efficiency. It is {\R4RS}.
It also has a module system designed by Jonathan Rees.
Scsh's design is not {\scm} specific, although the current implementation
is necessarily so.
Scsh is intended to be implementable in other {\Scheme} implementations.
The {\scm} virtual machine that scsh uses is a specially modified version;
standard {\scm} virtual machines cannot be used with the scsh heap image.
There are several different ways to invoke scsh.
You can run it as an interactive Scheme system, with a standard
read-eval-print interaction loop.
Scsh can also be invoked as the interpreter for a shell script by putting
a ``\verb|#!/usr/local/bin/scsh -s|'' line at the top of the shell script.
Descending a level, it is also possible to invoke the underlying virtual
machine byte-code interpreter directly on dumped heap images.
Scsh programs can be pre-compiled to byte-codes and dumped as raw,
binary heap images.
Writing heap images strips out unused portions of the scsh runtime
(such as the compiler, the debugger, and other complex subsystems),
reducing memory demands and saving loading and compilation times.
The heap image format allows for an initial \verb|#!/usr/local/lib/scsh/scshvm| trigger
on the first line of the image, making heap images directly executable as
another kind of shell script.
Finally, scsh's static linker system allows dumped heap images to be compiled
to a raw Unix a.out(5) format, which can be linked into the text section
of the vm binary.
This produces a true Unix executable binary file.
Since the byte codes comprising the program are in the file's text section,
they are not traced or copied by the garbage collector, do not occupy space
in the vm's heap, and do not need to be loaded and linked at startup time.
This reduces the program's startup time, memory requirements,
and paging overhead.
This chapter will cover these various ways of invoking scsh programs.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{Scsh command-line switches}
When the scsh top-level starts up, it scans the command line
for switches that control its behaviour.
These arguments are removed from the command line;
the remaining arguments can be accessed as the value of
the scsh variable \ex{command-line-arguments}.
\subsection{Scripts and programs}
The scsh command-line switches provide sophisticated support for
the authors of shell scripts and programs;
they also allow the programmer to write programs
that use the {\scm} module system.
There is a difference between a \emph{script}, which performs its action
\emph{as it is loaded}, and a \emph{program}, which is loaded/linked,
and then performs its action by having control transferred to an entry point
(\eg, the \ex{main()} function in C programs) that was defined by the
load/link operation.
A \emph{script}, by the above definition, cannot be compiled by the simple
mechanism of loading it into a scsh process and dumping out a heap image---it
executes as it loads. It does not have a top-level \ex{main()}-type entry
point.
It is more flexible and useful to implement a system
as a program than as a script.
Programs can be compiled straightforwardly;
they can also export procedural interfaces for use by other Scheme packages.
However, scsh supports both the script and the program style of programming.
\subsection{Inserting interpreter triggers into scsh programs}
When Unix tries to execute an executable file whose first 16 bits are
the character pair ``\ex{\#!}'', it treats the file not as machine-code
to be directly executed by the native processor, but as source code to
be executed by some interpreter.
The interpreter to use is specified immediately after the ``\ex{\#!}''
sequence on the first line of the source file
(along with one optional initial argument).
The kernel reads in the name of the interpreter, and executes that instead.
The interpreter is passed the source filename as its first argument, with
the original arguments following.
Consult the Unix man page for the \ex{exec} system call for more information.
Scsh allows Scheme programs to have these triggers placed on
their first line.
Scsh treats the character sequence ``\ex{\#!}'' as a block-comment sequence,%
\footnote{Why a block-comment instead of an end-of-line delimited comment?
See the section on meta-args.}
and skips all following characters until it reads the comment-terminating
sequence newline/exclamation-point/sharp-sign/newline (\ie, the
sequence ``\ex{!\#}'' occurring on its own line).
In this way, the programmer can arrange for an initial
\begin{code}
#!/usr/local/bin/scsh -s
!#\end{code}
header appearing in a Scheme program
to be ignored when the program is loaded into scsh.
\subsection{Module system}
Scsh uses the {\scm} module system, which defines
\emph{packages}, \emph{structures}, and \emph{interfaces}.
%
\begin{description}
\item [Package] A package is an environment---that is, a set of
variable/value bindings.
You can evaluate Scheme forms inside a package, or load a file into a package.
Packages export sets of bindings; these sets are called \emph{structures}.
\item [Structure] A structure is a named view on a package---a set of
bindings. Other packages can \emph{open} the structure, importing its
bindings into their environment. Packages can provide more than one
structure, revealing different portions of the package's environment.
\item [Interface] An interface is the ``type'' of a structure. An
interface is the set of names exported by a structure. These names
can also be marked with other static information (\eg, advisory type
declarations, or syntax information).
\end{description}
More information on the the {\scm} module system can be found in the
file \ex{module.ps} in the \ex{doc} directory of the {\scm} and scsh releases.
Programming Scheme with a module system is different from programming
in older Scheme implementations,
and the associated development problems are consequently different.
In Schemes that lack modular abstraction mechanisms,
everything is accessible; the major problem is preventing name-space conflicts.
In Scheme 48, name-space conflicts vanish; the major problem is that not
all bindings are accessible from every place.
It takes a little extra work to specify what packages export which values.
It may take you a little while to get used to the new style of program
development.
Although scsh can be used without referring to the module system at
all, we recommend taking the time to learn and use it.
The effort will pay off in the construction of modular, factorable programs.
\subsection{Switches}
\label{sec:scsh-switches}
The scsh top-level takes command-line switches in the following format:
%
\codex{scsh [\var{meta-arg}] [\vari{switch}i {\ldots}]
[\var{end-option} \vari{arg}1 {\ldots} \vari{arg}n]}
where
\begin{inset}
\begin{flushleft}
\begin{tabular}{ll@{\qquad}l}
\var{meta-arg:} & \verb|\| \var{script-file-name} \\
\\
\var{switch:} & \ex{-e} \var{entry-point}
& Specify top-level entry-point. \\
& \ex{-o} \var{structure}
& Open structure in current package. \\
& \ex{-m} \var{structure}
& Switch to package. \\
& \ex{-n} \var{new-package}
& Switch to new package. \\ \\
& \ex{-lm} \var{module-file-name}
& Load module into config package. \\
& \ex{-l} \var{file-name}
& Load file into current package. \\
& \ex{-dm} & Do script module. \\
& \ex{-ds} & Do script. \\
\\
\var{end-option:} & \ex{-s} \var{script} \\
& \ex{-sfd} \var{num} \\
& \ex{-c} \var{exp} \\
& \ex{--}
\end{tabular}
\end{flushleft}
\end{inset}
%
These command-line switches
essentially provide a little linker language for linking a shell script or a
program together with {\scm} modules.
The command-line processor serially opens structures and loads code into a
given package.
Switches that side-effect a package operate on a particular ``current''
package; there are switches to change this package.
(These switches provide functionality equivalent to the interactive
\ex{,open} \ex{,load} \ex{,in} and \ex{,new} commands.)
Except where indicated, switches specify actions that are executed in a
left-to-right order.
The initial current package is the user package, which is completely
empty and opens (imports the bindings of) the R4RS and scsh structures.
If the Scheme process is started up in an interactive mode, then the current
package in force at the end of switch scanning is the one inside which
the interactive read-eval-print loop is started.
The command-line switch processor works in two passes:
it first parses the switches, building a list of actions to perform,
then the actions are performed serially.
The switch list is terminated by one of the \var{end-option} switches.
The \vari{arg}{i} arguments occurring after an end-option switch are
passed to the scsh program as the value of \ex{command-line-arguments}
and the tail of the list returned by \ex{(command-line)}.
That is, an \var{end-option} switch separates switches that control
the scsh ``machine'' from the actual arguments being passed to the scsh
program that runs on that machine.
The following switches and end options are defined:
\begin{itemize}
\def\Item#1{\item{\ex{#1}}\\}
\Item{-o \var{struct}}
Open the structure in the current package.
\Item{-n \var{package}}
Make and enter a new package. The package has an associated structure
named \var{package} with an empty export list.
If \var{package} is the string ``\ex{\#f}'',
the new package is anonmyous, with no associated named structure.
The new package initially opens no other structures,
not even the R4RS bindings. You must follow a ``\ex{-n foo}''
switch with ``\ex{-o scheme}'' to access the standard identifiers such
as \ex{car} and \ex{define}.
\Item{-m \var{struct}}
Change the current package to the package underlying
structure \var{struct}.
(The \ex{-m} stands for ``module.'')
\Item{-lm \var{module-file-name}}
Load the specified file into scsh's config package --- the file
must contain source written in the Scheme 48 module language
(``load module''). Does not alter the current package.
\Item{-l \var{file-name}}
Load the specified file into the current package.
\Item{-c \var{exp}}
Evaluate expression \var{exp} in the current package and exit.
This is called \ex{-c} after a common shell convention (see sh and csh).
The expression is evaluated in the the current package (and hence is
affected by \ex{-m}'s and \ex{-n}'s.)
When the scsh top-level constructs the scsh command-line in this case,
it takes \ex{"scsh"} to be the program name.
This switch terminates argument scanning; following args become
the tail of the command-line list.
\Item{-e \var{entry-point}}
Specify an entry point for a program. The \var{entry-point} is
a variable that is taken from the current package in force at the end
of switch evaluation. The entry point does not have to be exported
by the package in a structure; it can be internal to the package.
The top level passes control to the entry point by applying it to
the command-line list (so programs executing in private
packages can reference their command-line arguments without opening
the \ex{scsh} package to access the \ex{(command-line)} procedure).
Note that, like the list returned by the \ex{(command-line)} procedure,
the list passed to the entry point includes the name
of the program being executed (as the first element of the list),
not just the arguments to the program.
A \ex{-e} switch can occur anywhere in the switch list, but it is the
\emph{last} action performed by switch scanning if it occurs.
(We violate ordering here as the shell-script \ex{\#!} mechanism
prevents you from putting the \emph{-e} switch last, where it belongs.)
\Item{-s \var{script}}
Specify a file to load.
A \ex{-ds} (do-script) or \ex{-dm} (do-module) switch occurring earlier in
the switch list gives the place where the script should be loaded. If
there is no \ex{-ds} or \ex{-dm} switch, then the script is loaded at the
end of switch scanning, into the module that is current at the end of
switch scanning.
We use the \ex{-ds} switch to violate left-to-right switch execution order
as the \ex{-s} switch is \emph{required} to be last
(because of the \ex{\#!} machinery),
independent of when/where in the switch-processing order
it should be loaded.
When the scsh top-level constructs the scsh command-line in this case,
it takes \var{script} to be the program name.
This switch terminates switch parsing; following args are ignored
by the switch-scanner and are passed through to the program as
the tail of the command-line list.
\Item{-sfd \var{num}}
Loads the script from file descriptor \var{num}.
This switch is like the \ex{-s} switch,
except that the script is loaded from one of the process' open input
file descriptors.
For example, to have the script loaded from standard input, specify
\ex{-sfd 0}.
\Item{--}
Terminate argument scanning and start up scsh in interactive mode.
If the argument list just runs out, without either a terminating
\ex{-s} or \ex{--} arg, then scsh also starts up in interactive mode,
with an empty \ex{command-line-arguments} list
(for example, simply entering \ex{scsh} at a shell prompt with no
args at all).
When the scsh top-level constructs the scsh command-line in this case,
it takes \ex{"scsh"} to be the program name.
This switch terminates switch parsing; following args are ignored
by the switch-scanner and are passed through to the program as
the tail of the command-line list.
\Item{-ds}
Specify when to load the script (``do-script''). If this switch occurs,
the switch list \emph{must} be terminated by a \ex{-s \var{script}}
switch. The script is loaded into the package that is current at the
\ex{-ds} switch.
\Item{-dm}
As above, but the current module is ignored. The script is loaded into the
\ex{config} package (``do-module''), and hence must be written in the
{\scm} module language.
This switch doesn't affect the current module---after executing this
switch, the current module is the same as as it was before.
This switch is provided to make it easy to write shell scripts in the
{\scm} module language.
\end{itemize}
\subsection{The meta argument}
\label{sec:meta-arg}
The scsh switch parser takes a special command-line switch,
a single backslash called the ``meta-argument,'' which is useful for
shell scripts.
If the initial command-line argument is a ``\verb|\|''
argument, followed by a filename argument \var{fname}, scsh will open the file
\var{fname} and read more arguments from the second line of this file.
This list of arguments will then replace the ``\verb|\|'' argument---\ie,
the new arguments are inserted in front of \var{fname},
and the argument parser resumes argument scanning.
This is used to overcome a limitation of the \ex{\#!} feature:
the \ex{\#!} line can only specify a single argument after the interpreter.
For example, we might hope the following scsh script, \ex{ekko},
would implement a simple-minded version of the Unix \ex{echo} program:
\begin{code}
#!/usr/local/bin/scsh -e main -s
!#
(define (main args)
(map (\l{arg} (display arg) (display " "))
(cdr args))
(newline))\end{code}
%
The idea would be that the command
\codex{ekko Hi there.}
would by expanded by the \ex{exec(2)} kernel call into
%
\begin{code}
/usr/local/bin/scsh -e main -s ekko Hi there.\end{code}
%
In theory, this would cause scsh to start up, load in file \ex{ekko},
call the entry point on the command-line list
\codex{(main '("ekko" "Hi" "there."))}
and exit.
Unfortunately, the {\Unix} \ex{exec(2)} syscall's support for scripts is
not very general or well-designed.
It will not handle multiple arguments;
the \ex{\#!} line is usually required to contain no more than 32 characters;
it is not recursive.
If these restrictions are violated, most Unix systems will not provide accurate
error reporting, but either fail silently, or simply incorrectly implement
the desired functionality.
These are the facts of Unix life.
In the \ex{ekko} example above, our \ex{\#!} trigger line has three
arguments (``\ex{-e}'', ``\ex{main}'', and ``\ex{-s}''), so it will not
work.
The meta-argument is how we work around this problem.
We must instead invoke the scsh interpreter with the single \cd{\\} argument,
and put the rest of the arguments on line two of the program.
Here's the correct program:
%
\begin{code}
#!/usr/local/bin/scsh \\
-e main -s
!#
(define (main args)
(map (\l{arg} (display arg) (display " "))
(cdr args))
(newline))\end{code}
%
Now, the invocation starts as
\codex{ekko Hi there.}
and is expanded by exec(2) into
\begin{code}
/usr/local/bin/scsh \\ ekko Hi there.\end{code}
When scsh starts up, it expands the ``\cd{\\}'' argument into the arguments
read from line two of \ex{ekko}, producing this argument list:
\begin{code}\cddollar
\underline{-e main -s ekko} Hi there.
$\uparrow$
{\rm{}Expanded from} \cd{\\} ekko\end{code}
%
With this argument list, processing proceeds as we intended.
\subsubsection{Secondary argument syntax}
Scsh uses a very simple grammar to encode the extra arguments on
the second line of the scsh script.
The only special characters are space, tab, newline, and backslash.
\begin{itemize}
\item Each space character terminates an argument.
This means that two spaces in a row introduce an empty-string argument.
\item The tab character is not permitted
(unless you quote it with the backslash character described below).
This is to prevent the insidious bug where you believe you have
six space characters, but you really have a tab character,
and \emph{vice-versa}.
\item The newline character terminates an argument, like the space character,
and also terminates the argument sequence.
This means that an empty line parses to the singleton list whose one
element is the empty string: \ex{("")}.
The grammar doesn't admit the empty list.
\item The backslash character is the escape character.
It escapes backslash, space, tab, and newline, turning off their
special functions, and allowing them to be included in arguments.
The {\Ansi} C escape sequences (\verb|\b|, \verb|\n|, \verb|\r|
and \verb|\t|) are also supported;
these also produce argument-constituents---\verb|\n| doesn't act
like a terminating newline.
The escape sequence \verb|\|\emph{nnn} for \emph{exactly} three
octal digits reads as the character whose {\Ascii} code is \emph{nnn}.
It is an error if backslash is followed by just one or two octal digits:
\verb|\3Q| is an error.
Octal escapes are always constituent chars.
Backslash followed by other chars is not allowed
(so we can extend the escape-code space later if we like).
\end{itemize}
You have to construct these line-two argument lines carefully.
In particular, beware of trailing spaces at the end of the line---they'll
give you extra trailing empty-string arguments.
Here's an example:
%
\begin{inset}
\begin{verbatim}
#!/bin/interpreter \
foo bar quux\ yow\end{verbatim}
\end{inset}
%
would produce the arguments
%
\codex{("foo" "bar" "" "quux yow")}
\subsection{Examples}
\begin{itemize}
\def\Item#1{\item{\ex{#1}}\\}
\def\progItem#1{\item{Program \ex{#1}}\\}
\Item{scsh -dm -m myprog -e top -s myprog.scm}
Load \ex{myprog.scm} into the \ex{config} package, then shift to the
\ex{myprog} package and call \ex{(top '("myprog.scm"))}, then exit.
This sort of invocation is typically used in \ex{\#!} script lines
(see below).
\Item{scsh -c '(display "Hello, world.")'}
A simple program.
\Item{scsh -o bigscheme}
Start up interactively in the user package after opening
structure \ex{bigscheme}.
\Item{scsh -o bigscheme -- Three args passed}
Start up interactively in the user package after opening \ex{bigscheme}.
The \ex{command-line-args} variable in the scsh package is bound to the
list \ex{("Three" "args" "passed")}, and the \ex{(command-line)}
procedure returns the list \ex{("scsh" "Three" "args" "passed")}.
\progItem{ekko}
This shell script, called \ex{ekko}, implements a version of
the Unix \ex{echo} program:
\begin{code}
#!/usr/local/bin/scsh -s
!#
(for-each (\l{arg} (display arg) (display " "))
command-line-args)\end{code}
Note this short program is an example of a \emph{script}---it
executes as it loads.
The Unix rule for executing \ex{\#!} shell scripts causes
\codex{ekko Hello, world.}
to expand as
\codex{/usr/local/bin/scsh -s ekko Hello, world.}
\progItem{ekko}
This is the same program, \emph{not} as a script.
Writing it this way makes it possible to compile the program
(and then, for instance, dump it out as a heap image).
%
\begin{code}
#!/usr/local/bin/scsh \\
-e top -s
!#
(define (top args)
(for-each (\l{arg} (display arg) (display " "))
(cdr args)))\end{code}
%
The \ex{exec(2)} expansion of the \ex{\#!} line together with
the scsh expansion of the ``\verb|\ ekko|'' meta-argument
(see section~\ref{sec:meta-arg}) gives the following command-line expansion:
\begin{code}
ekko Hello, world.
{\evalto} /usr/local/bin/scsh \\ ekko Hello, world.
{\evalto} /usr/local/bin/scsh -e top -s ekko Hello, world.\end{code}
\progItem{sort}
This is a program to replace the Unix \ex{sort} utility---sorting lines
read from stdin, and printing the results on stdout.
Note that the source code defines a general sorting package,
which is useful (1) as a Scheme module exporting sort procedures
to other Scheme code, and (2) as a standalone program invoked from
the \ex{top} procedure.
\begin{code}
#!/usr/local/bin/scsh \\
-dm -m sort-toplevel -e top -s
!#
;;; This is a sorting module. TOP procedure exports
;;; the functionality as a Unix program akin to sort(1).
(define-structures ((sort-struct (export sort-list
sort-vector!))
(sort-toplevel (export top)))
(open scheme)
(begin (define (sort-list elts <=) {\ldots})
(define (sort-vec! vec <=) {\ldots})
;; Parse the command line and
;; sort stdin to stdout.
(define (top args)
{\ldots})))\end{code}
The expansion below shows how the command-line scanner
(1) loads the config file \ex{sort} (written in the {\scm} module language),
(2) switches to the package underlying the \ex{sort-toplevel} structure,
(3) calls \ex{(top '("sort" "foo" "bar"))} in the package, and finally
(4) exits.
%
{\small
\begin{centercode}
sort foo bar
{\evalto} /usr/local/bin/scsh \\ sort foo bar
{\evalto} /usr/local/bin/scsh -dm -m sort-toplevel -e top -s sort foo bar\end{centercode}}
An alternate method would have used a
\begin{code}
-n #f -o sort-toplevel\end{code}
sequence of switches to specify a top-level package.
\end{itemize}
Note that the sort example can be compiled into a Unix program by
loading the file into an scsh process, and dumping a heap with top-level
\ex{top}. Even if we don't want to export the sort's functionality as a
subroutine library, it is still useful to write the sort program with the
module language. The command line design allows us to run this program as
either an interpreted script (given the \ex{\#!} args in the header) or as a
compiled heap image.
\subsection{Process exit values}
Scsh ignores the value produced by its top-level computation when determining
its exit status code.
If the top-level computation completed with no errors,
scsh dies with exit code 0.
For example, a scsh process whose top-level is specified by a \ex{-c \var{exp}}
or a \ex{-e \var{entry}} entry point ignores the value produced
by evaluating \var{exp} and calling \var{entry}, respectively.
If these computations terminate with no errors, the scsh process
exits with an exit code of 0.
To return a specific exit status, use the \ex{exit} procedure explicitly, \eg,
\begin{tightcode}
scsh -c \\
"(exit (status:exit-val (run (| (fmt) (mail shivers)))))"\end{tightcode}
\section{The scsh virtual machine}
To run the {\scm} implementation of scsh, you run a specially modified
copy of the {\scm} virtual machine with a scsh heap image.
The scsh binary is actually nothing but a small cover program that invokes the
byte-code interpreter on the scsh heap image for you.
This allows you to simply start up an interactive scsh from a command
line, as well as write shell scripts that begin with the simple trigger
\codex{\#!/usr/local/bin/scsh -s}
You can also directly execute the virtual machine,
which takes its own set of command-line switches..
For example,
this command starts the vm up with a 1Mword heap (split into two semispaces):
\codex{scshvm -o scshvm -h 1000000 -i scsh.image arg1 arg2 \ldots}
The vm peels off initial vm arguments
up to the \ex{-i} heap image argument, which terminates vm argument parsing.
The rest of the arguments are passed off to the scsh top-level.
Scsh's top-level removes scsh switches, as discussed in the previous section;
the rest show up as the value of \ex{command-line-arguments}.
Directly executing the vm can be useful to specify non-standard switches, or
invoke the virtual machine on special heap images, which can contain
pre-compiled scsh programs with their own top-level procedures.
\subsection{VM arguments}
\label{sec:vm-args}
The vm takes arguments in the following form:
\codex{scshvm [\var{meta-arg}] [\var{vm-options}\+] [\var{end-option} \var{scheme-args}]}
where
\begin{inset}
\begin{tabular}{ll}
\var{meta-arg:} & \verb|\ |\var{filename} \\
\\
\var{vm-option}: & \ex{-h }\var{heap-size-in-words} \\
& \ex{-s }\var{stack-size-in-words} \\
& \ex{-o }\var{object-file-name} \\
\\
\var{end-option:} & \ex{-i }\var{image-file-name} \\
& \ex{--}
\end{tabular}
\end{inset}
The vm's meta-switch ``\verb|\ |\var{filename}'' is handled the same
as scsh's meta-switch, and serves the same purpose.
\subsubsection{VM options}
The \ex{-o \var{object-file-name}} switch tells the vm where to find
relocation information for its foreign-function calls.
Scsh will use a pre-compiled default if it is not specified.
Scsh \emph{must} have this information to run,
since scsh's syscall interfaces are done with foreign-function calls.
The \ex{-h} and \ex{-s} options tell the vm how much space to allocate
for the heap and stack.
The heap size value is the total number of words allocated for the heap;
this space is then split into two semi-spaces for {\scm}'s stop-and-copy
collector.
\subsubsection{End options}
End options terminate argument parsing.
The \ex{-i} switch is followed by the name of a heap image for the
vm to execute.
The \var{image-file-name} string is also taken to be the name of the program
being executed by the VM; this name becomes the head of the argument
list passed to the heap image's top-level entry point.
The tail of the argument list is constructed from all following arguments.
The \ex{--} switch terminates argument parsing without giving
a specific heap image; the vm will start up using a default
heap (whose location is compiled into the vm).
All the following arguments comprise the tail of the list passed off to
the heap image's top-level procedure.
Notice that you are not allowed to pass arguments to the heap image's
top-level procedure (\eg, scsh) without delimiting them with \ex{-i}
or \ex{--} flags.
\subsection{Inserting interpreter triggers into heap images}
{\scm}'s heap image format allows for an informational header:
when the vm loads in a heap image, it ignores all data occurring before
the first control-L character (\textsc{Ascii} 12).
This means that you can insert a ``\ex{\#!}'' trigger line into a
heap image, making it a form of executable ``shell script.''
Since the vm requires multiple arguments to be given on the command
line, you must use the meta-switch.
Here's an example heap-image header:
\begin{code}
#!/usr/local/lib/scsh/scshvm \\
-o /usr/local/lib/scsh/scshvm -i
{\ldots} \textnormal{\emph{Your heap image goes here}} \ldots\end{code}
\subsection{Inserting a double-level trigger into Scheme programs}
If you're a nerd, you may enjoy doing a double-level machine shift
in the trigger line of your Scheme programs with the following magic:
\begin{code}\small
#!/usr/local/lib/scsh/scshvm \\
-o /usr/local/lib/scsh/scshvm -i /usr/local/lib/scsh/scsh.image -s
!#
{\ldots} \textnormal{\emph{Your Scheme program goes here}} \ldots\end{code}
\section{Compiling scsh programs}
Scsh allows you to create a heap image with your own top-level procedure.
Adding the pair of lines
\begin{code}
#!/usr/local/lib/scsh/scshvm \\
-o /usr/local/lib/scsh/scshvm -i\end{code}
to the top of the heap image will turn it into an executable {\Unix} file.
You can create heap images with the following two procedures.
\defun{dump-scsh-program}{main fname}{\undefined}
\begin{desc}
This procedure writes out a scsh heap image. When the
heap image is executed by the {\scm} vm, it will call
the \var{main} procedure, passing it the vm's argument list.
When \ex{main} returns an integer value $i$, the vm exits with
exit status $i$.
The {\Scheme} vm will parse command-line switches as
described in section~\ref{sec:vm-args}; remaining arguments
form the tail of the command-line list that is passed to \ex{main}.
(The head of the list is the name of the program being executed
by the vm.)
Further argument parsing
(as described for scsh in section~\ref{sec:scsh-switches})
is not performed.
The heap image created by \ex{dump-scsh-program} has unused
code and data pruned out, so small programs compile to much smaller
heap images.
\end{desc}
\defun{dump-scsh}{fname}{\undefined}
\begin{desc}
This procedure writes out a heap image with the standard
scsh top-level.
When the image is resumed by the vm, it will parse and
execute scsh command-line switches as described in section
\ref{sec:scsh-switches}.
You can use this procedure to write out custom scsh heap images
that have specific packages preloaded and start up in specific
packages.
\end{desc}
Unfortunately, {\scm} does not support separate compilation of
Scheme files or Scheme modules.
The only way to compile is to load source and then dump out a
heap image.
One occasionally hears rumours that this is being addressed
by the {\scm} development team.
\section{Statically linking heap images}
The static heap linker converts a {\scm} bytecode image contained
in a .image file to a C representation. This C code is then compiled and
linked in with a virtual machine, producing a single executable.
Some of the benefits are:
\begin{itemize}
\item Instantaneous start-up time.
\item Improved paging; scsh images can be shared between different
processes.
\item Vastly reduced GC copying---the whole initial image
is moved out of the heap, and neither traced nor copied.
\item Result program no longer depends on the filesystem for its
initial image.
\end{itemize}
The static heap linker takes arguments in the following form:
\codex{scsh-hlink \var{image} \var{executable} [\var{option} \ldots]}
It reads in the heap image \var{image}, translates it into C code,
compiles the C code, and links it against the scsh vm, producing the
standalone binary file \var{executable}.
Each C file represents part of the heap image as a constant C \ex{long} vector
that looks something like this:
{\small\begin{verbatim}
const long p116[]={0x882,0x24,0x19,
0x882,(long)(&p19[785])+7,(long)(&p119[125])+7,
0x882,(long)(&p119[128])+7,(long)(&p119[131])+7,
0x882,(long)(&p102[348])+7,(long)(&p3[114])+7,
0xfc2,0x2030200,0x7100209,0x1091002,0x1c075a,
0x882,(long)(&p29[1562])+7,(long)(&p119[137])+7,
0x882,(long)(&p78[692])+7,(long)(&p119[140])+7,
.
.
.
};
\end{verbatim}}%
%
Translating to a C declaration gives us freedom from the various
object-file formats.\footnote{This idea is due to Jonathan Rees.}
Note that the const declaration allows the compiler to put this array in the
text pages of the executable.
The heap is split into parts because many C compilers cannot handle
multi-megabyte initialised vector declarations.
The allowed options to the heap linker are:
\begin{itemize}
\def\Item#1{\item{\ex{#1}}\\}
\Item{--temp \var{dir}} The temporary directory to hold .c and .o files.
The default is typically configured to be
\ex{/usr/tmp}, and can be overridden by the
environment variable \ex{TMPDIR}.
Example:
\codex{--temp /tmp}
\Item{--cc \var{command}} The command to run the C compiler.
The default can be overridden by the environment
variable \ex{CC}.
Example:
\codex{--cc "gcc -g -O"}
\Item{--ld \var{command}} The arguments to run the C compiler as a linker.
The default can be overridden by the
environment variable \ex{LDFLAGS}.
Example:
\codex{--ld "-Wl,-E"}
\Item{--libs \var{libs}} The libraries needed to link the VM and heap.
The default can be overridden by the
environment variable \ex{LIBS}.
Example:
\codex{--libs "-ldld -lld -lm"}
\end{itemize}
Be warned that the current heap linker has many shortcomings.
\begin{itemize}
\item It is extremely slow. Really, really slow. Translating the standard
scsh heap image into a standalone binary takes well over an hour on a
40Mb/133Mhz Pentium system.
A memory-starved 486 could take all night.
\item It cannot be applied to itself. The current implementation
works by replacing some of the heap-dumping code. This means
you cannot load the heap-linker code into a scsh system and
subsequently use \ex{dump-scsh-program} to create a heap-linker
heap image.
\item The interface leaves a lot to be desired.
\begin{itemize}
\item It requires the heap image to be referenced by a file-name;
the linker will not allow you to feed it the input heap image
on a port.
\item The heap-image is linked against the vm contained in
\begin{tightcode}
/usr/local/lib/scsh/libscshvm.a\end{tightcode}
This is wired in at the time scsh is installed on your system.
\item There is no Scheme procedural interface.
\end{itemize}
\item The program produced uses the default VM argv parser \verb|process_args|
from the scsh source file \ex{main.c} to process the command line
before handing it off to the heap image's top-level procedure.
This is not what you want for many programs.
The system needs to be changed to allow users to override this default
with their own VM argument parsers.
\item A possible problem is the Unix limits on the number of command
line arguments. The heap-linker calls the C linker with a large number of
object files. Its conceivable that on some Unix systems this could fail
now or if scsh grows in the future. The solution could be to create
library archives of a few dozen files and then link the result few dozen
library archives to make the executable.
\end{itemize}
In spite of these many shortcomings, we are providing the static linker
as it stands in this release so that people may get some experience with
it.
Here is an example of how one might use the heap linker:
\begin{code}
scsh-hlink scsh.image fastscsh\end{code}
We'd love it if someone would dive into the source and improve it.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{Standard file locations}
Because the scshvm binary is intended to be used for writing shell
scripts, it is important that the binary be installed in a standard
place, so that shell scripts can dependably refer to it.
The standard directory for the scsh tree should be \ex{/usr/local/lib/scsh/}.
Whenever possible, the vm should be located in
\codex{/usr/local/lib/scsh/scshvm}
and a scsh heap image should be located in
\codex{/usr/local/lib/scsh/scsh.image}
The top-level scsh program should be located in
\codex{/usr/local/lib/scsh/scsh}
with a symbolic link to it from
\codex{/usr/local/bin/scsh}
The {\scm} image format allows heap images to have \ex{\#!} triggers,
so \ex{scsh.image} should have a \ex{\#!} trigger of the following form:
\begin{code}
#!/usr/local/lib/scsh/scshvm \\
-o /usr/local/lib/scsh/scshvm -i
{\ldots} \textnormal{\emph{heap image goes here}} \ldots\end{code}
Reference in New Issue View Git Blame Copy Permalink