Languguage OS 2/Languguage OS II Version 10-94 (Knowledge Media)(1994).ISO/gnu/objcissu.lha/exception-handling

Text File | 1993-03-01 | 26KB | 605 lines

Paul Murphy <[e-mail]> just forwarded this to me.  FYI.  10/14/92 gsk

Position Paper
ECOOP'91 WORKSHOP ON
EXCEPTION HANDLING AND OBJECT-ORIENTED PROGRAMMING
July 1991
Geneva, Switzerland

Brad Cox
203 426 1875
[e-mail]
CI$ 71230,647
The Stepstone Corporation
75 Glen Road
Sandy Hook CT 06482

This consists of selected sections from a much longer paper,
`TaskMaster', which discusses a number of environmental features,
including exception handling and lightweight multi-tasking, that can
be supported in any C-based language with a single uniform language
extension called `action expressions'.

The first section of this document discusses exception handling as a
topic unto itself. The final section introduces action expression
syntax and semantics, as is being considered for adoption as an
extension to the Objective-C programming environment.

Exception handling

Consider a subroutine, main(), which calls foo(), which calls bar().
The runtime stack that underlies the C runtime environment extends to
record that main has called foo and that foo has called bar. Then it
retracts as bar returns to foo and as foo returns to main. The same
call/return path is used unconditionally, regardless of whether the
subroutines ran successfully or failed because of an exception.

The absence of a way of handling exceptions explicitly, independently
from normal processing, is a severe obstacle to software quality and
reusability. Since subroutines routines return the same way,
regardless of whether they succeed or fail, a method that should
return a handle to a new object might instead return an error code to
indicate that it could not, perhaps because of insufficient memory.
Since any such call might fail, the caller must check every return
value, reporting any failures to higher levels with similar means.
This is sufficiently tedious that it is neglected, allowing unhandled
exceptions to crash the application.

An `exception' is a situation where a computation does not proceed as
planned, perhaps because of I/O errors, inability to allocate a new
object because of insufficient memory, or defects in the software
itself. `Exception handling' is a language/environmental feature that
reserves the normal subroutine/message return channel exclusively for
normal processing. Routines that return normally can be assumed to
have succeeded, since those that fail will return via a independent
channel reserved for exceptions. Low-level routines never fail by
returning an error code for the caller to decipher, but by `raising an
exception'. Higher level routines establish code fragments, `exception
handlers', that will receive control if the exception they are to
handle is raised by a lower-level routine.

The following shows how exception handling might be done in C, in
order to show the limitations of this solution and how these
limitations are addressed in TaskMaster. TRY() is a C macro that uses
setjmp() to record the machine's register settings before entering the
computation that might fail; the foo subroutine.

main() {
    TRY() {
        int v = foo();
        <normal processing<
    } HANDLE(OUTOFMEMORYEXCEPTION) {
        <handle low memory exceptions<
    } HANDLE(IOEXCEPTION) {
        <handle IO failure exceptions<
    } OTHERWISE() {
        <handle any other exceptions<
    }
}
foo() {
    TRY() {
        int v = bar();
        <normal processing<
    } HANDLE(OUTOFMEMORYEXCEPTION);
}
bar() {
    id newObject = [SomeClass new];
    <normal processing<
}

If the foo() subroutine, or any subroutine that it calls, fails
because of some exception, it does not return normally. Instead, it
`raises an exception', which returns directly to the higher-level
routine, by using longjmp() to return directly, bypassing the
intervening subroutines altogether. The TRY() macro detects the
abnormal return and transfers to the appropriate handler, which
handles the exception in some case-specific fashion. By having the
TRY() macro manage saved register settings on LIFO stack, exception
handlers can be nested so that exceptions are presented to their
handlers in deepest-first sequence. The low-level exception handlers
may choose to pass control to the next higher-level handler by popping
the higher-level registers from the exception handler's stack and
calling longjmp() again.

The problems with this scheme will be familiar to anyone who has used
it extensively. Since the exception handlers unconditionally execute
in the context of the calling routine, as opposed to as a subroutine
of the routine that detected the exception, information is
unconditionally lost as to what sequence of calls triggered the
exception. For example, foo's OUTOFMEMORYEXCEPTION handler can
determine that bar, or one of its subroutines, failed because of
insufficient memory. But it cannot tell which one, because the stack
has been rewound before the handler has been invoked. This makes it
useless for exception handlers to invoke process inspectors
(debuggers), since the debugging information has been unconditionally
discarded.

TaskMaster takes a different approach. It exports the handler to the
exception, rather than the other way around. By invoking the exception
handler as a subroutine of the exception, the stack is never erased
until the user's code has had a chance to use it. One way to think
about the difference is that TaskMaster exception handlers are first
class C subroutines, not compound statements as in the preceding
example. They are passed as function pointers in the exception handler
stack and invoked as subroutines by the logic for raising an
exception.

Action expressions are a compiler-supported enhancement to deal with
the fact that code that uses function pointers extensively can be hard
to write and nearly impossible to understand later. This extension,
which is planned for a future compiler release and is not yet
available, would let this example be coded like this:

main() {
    [{ int v = foo();
        <normal processing<
    } ifException:{
        if (OUTOFMEMORYEXCEPTION) {
        <handle low memory exceptions<
        } else if (IOEXCEPTION) {
        <handle IO failure exceptions<
        } else {
        <handle any other exceptions<
        }
    }];
foo() {
    [{ int v = bar();
        <normal processing<
    } ifException: {
        <handle exceptions<
    }];
}
bar() {
    id newObject = [SomeClass new];
    <normal processing<
}

The statements that look like C compound statements used as
expressions, in {bold braces}, are action expressions, which will be
described in the next section. Their relevance to exception handling
is that

The code to be protected from exceptions, and the code for providing
that protection, can be written alongside each other in a familiar
readable fashion, without the syntactic clutter of working with
function pointers directly.

The compiler will generate the necessary boiler plate automatically,
transforming the action expressions into function definitions that the
underlying C compiler knows how to handle.  Although actions look
syntactically like expressions, they are semantically analogous to
function pointers that the exception handler can invoke as a
subroutine of the exception, as opposed to within the scope of the
calling site.

TaskMaster's run-time support for this style of exception handling is
in place today, and can be used by following the code generation
strategy that the extended compiler will use; coding action
instantiation statements by hand.

A final TaskMaster feature is that exception handling is closely
integrated with lightweight multitasking. Each task maintains an
independent stack of exception handlers that inherits any exception
handlers of its parent task. For example, a parent task can provide
supply handlers of last resort for exceptions in its subtasks.
TaskMaster automatically initializes the root task with a default
handler for all possible exceptions.

This handler of last resort features an interactive task inspector
with interactive debugging facilities. A programmer can use the
inspector to examine the call history that led to the exception and
specify whether to exit the application, produce a file suitable for
detailed debugging, or to continue in spite of the exception.

The inspector is based on a library of routines that interpret C call
histories in terms that a programmer can easily understand. For
example, call histories are not presented as hexadecimal numbers but
in symbolic terms. Subroutine and selector names are spelled out with
all arguments decoded. Object references are presented in terms of the
object's class name and its address, string pointers are presented in
terms of the string's contents, and so forth. These decoding
facilities are based on a library of platform-dependent routines in
the TaskMaster PDL. These routines support such platform-dependent
operations as loading the application's symbol table. They also
provide the heuristic rules that determine which format should be used
in presenting the otherwise untyped numbers in a C call stack.

Apart from such convenience features, the exception handling is not
particularly dependent on the target platform. Exception handling is
based on a pair of routines, setjmp() and longjmp(), in the standard C
run time library. Nor is it particularly dependent on multi-tasking,
except that bundling exception handling with multi-tasking makes it
possible for subtasks to manage exceptions independently from other
tasks. Nor is exception handling specific to Objective-C. Its benefits
are equally germane to ordinary C programs. This is another reason why
TaskMaster's exception handling facilities are brought forth through
two API's, a message-based API for users who want to treat exceptions
as objects and a function-based API for those who want to treat them
as functions.

Action expressions

Action expressions in Objective-C are related to block expressions in
Smalltalk, with differences that adapt to the constraints of
stack-based languages like C. Action expressions are a way to express
actions, or deferred computations; computations to be written at one
place but invoked from another.

The classical examples of a deferred computation is a computation that
a collection is to perform on each of its members or that a menu is to
perform when selected. As the following examples show, action
expressions are generally useful, especially for defining menus,
operating on collection members, handling exceptions, and defining
lightweight tasks:

[ aMenu str:"Open" action:{ code to do a menu operation }];
[ { code that might fail } ifException: { code to handle the exception } ];
[ anyCollection do:{ code to be applied to each member } ];
[ { code to run as a separate task } forkWith:2, arg1, arg2];

Here is an a simpler example of how action expressions help to express
a deferred computation. In this example, a printf() statement is to be
passed to a subroutine, bar(), which will execute it at its
convenience. Notice that the deferred computation is written right
inside the foo subroutine, just like any C expression. But it is not
to be executed there. The action is passed as an argument to the bar
routine, so that bar can decide whether, and when, to invoke it later.

extern int aGlobal;
foo(aFormal)
    int aLocal;

    <long, involved computation<

    /*
     * Pass this action to bar, which may run it later
     */
    bar({ int formalArg; | printf(<, aGlobal, aFormal, aLocal, formalArg);});

    <long, involved computation<
}

An action expression can be thought of as any legal C compound
statement, used in a context where an expression is otherwise
expected:

Action expressions are written in-line, exactly like an ordinary C or
Objective-C expression. In this case, the action expression is an
argument to the foo subroutine call.

Action expressions can freely reference any and all data within the
expression's enclosing scope, such as the aGlobal, aFormal, and aLocal
variables in this example.

Evaluation of an action expression produces an object; an instance of
class Action. For example, the foo subroutine will receive an instance
of class Action as its formal argument, anAction.

Invocation of an action occurs separately, just as invocation of a
function via a function pointer is separate from the act of defining
the function that it points to. The bar routine might invoke the
printf statement with the message expression, [anAction value], in
which formalArg will be nil. If it invokes it as [anAction
value:aValue], formalArg will contain aValue.

Action expressions can receive formal arguments from the message that
invokes them, such as formalArg in this example. These are analogous
to subroutine arguments. If formal arguments are to be used, they are
declared in the block expression as in this example, where the | is a
place-holder for an as-yet-undecided syntactic delimiter.

Action expression's usefulness becomes most obvious in contrast with
how this example would be coded in ordinary C, using function
pointers:

extern aGlobal;
static dummyFv, dummyLv;
static dummyFn(formalArg)
{
    printf("<", aGlobal, dummyFv, dummyLv, formalArg);
}
foo(aFormal) {
    int aLocal;

    <long, involved computation<

    /*
     * Pass dummyFn to bar as a function pointer
     * so that bar can run it later
     */
    dummyFv = aFormal;
    dummyLv = aLocal;
    bar(dummyFn);

    <long involved computation<
}

This is less desirable than the original code for reasons referred to
earlier as syntactic clutter.

The computation is no longer in-line, inside the foo routine. The
reader must be extremely careful to see that this code is packaging a
printf() statement for execution by bar().

The enclosing scope (foo) cannot pass data to the computation in a
clean, readable fashion. It must use extraneous global variables, the
dummyFv and dummyLv symbols in this example, to export data from its
internal scope to the external function, dummyFn. These global
variables are an obstacle to recursion and a rich opportunity for hard
to debug race conditions should foo be used by multiple tasks.

Objective-C actions have a key restriction that is not present with
Smalltalk blocks. The code inside a action accesses copies of the
variables that it references in the enclosing scope. These copies are
made when the action is instantiated, not when it is invoked, as is
the case with Smalltalk Blocks. The copies are stored as indexed
instance variables inside the Action object and not in the enclosing
scope (foo's stack frame).

When the underlying C function is invoked, these variables are
exported to the function via a structure pointer, as can be seen in
the following example. For example, the aGlobal, aFormal, and aLocal
variables in this example are copied when the action is created (i.e.
just before bar is called), not when the deferred computation is
invoked (inside bar). Objective-C action expressions do not share
memory with their enclosing scope, they cannot export changes to the
enclosing scope as in Smalltalk. Changes to action variables is
ineffective, since the change affects a copy, not the original.

This may have made the simple seem unduly complicated. What is going
on here can seen by examining the C code that the compiler would emit
for this example:

static void fn(struct { int aGlobal, aFormal, aLocal; } *c, int formalArg)
{
    printf(<, c->aGlobal, c->aFormal, c->aLocal, formalArg);
}
foo(aFormal) {
    extern int aGlobal;
    int aLocal;

    <long, involved computation<

    bar(createAction(fn, 3, aGlobal, aFormal, aLocal));

    <long, involved computation<

}

Of course, this example was intentionally simplified by choosing an
example in which everything is of type int. This does not mean that
actions will only be usable for integers; just that the adjustments
that the compiler would emit for other types seem obvious. Until the
compiler has been enhanced to emit this code automatically, this
example shows what a present-day Objective-C user would write to use
TaskMaster's action-based library facilities today.


Date: Mon, 12 Oct 1992 13:46:03 -0500
To: [e-mail]
From: [e-mail] (Brad Cox)
Subject: Action Expressions (exception handling, multitasking)
Cc: Bruce Nilo <[e-mail]>

In Re: Your mission, should you choose to accept it. . Bruce Nilo
<[e-mail]> writes

NeXT has half heartedly implemented an "Exception Handling" facility.
It basically consists of some macros, and the use of setjmp() and
longjmp(). An error handling facility should be specified,
implemented, and adhered to throughout the runtime system.

I agree wholeheartedly! I joined this list primarily to encourage
extensions to Objective-C of precisely this nature.

I made substantial progress on the run-time part of this problem at Stepstone
several years ago, but never managed to complete the language extensions
(upwards compatible) to make them broadly useful.

<Taskmaster Document> "Action expressions in Objective-C are similar to
block expressions in Smalltalk, with differences to adapt to the
constraints of stack-based languages like C. Action expressions are a way
to express actions, or deferred computations; computations to be written at
one place but invoked from another."

This document, which describes the (implemented) runtime components and
(unimplemented) syntactic extensions, would be useful in considering
language and/or runtime extensions.

I'll be glad to email a copy to those planning to  work on such extentions.
(Please...only if you're planning to do actual work!)
===
Brad Cox, Ph.D; Program on Social and Organizational Learning; George Mason
University; Fairfax VA 22030; 703 691 3187 direct; 703 993 1142 reception;
[e-mail]
---
Information Age Consulting; 13668 Bent Tree Circle #203; Centreville VA
22020;           703 968 8229 home; 703 968 8798 fax; [e-mail]

Date: Thu, 15 Oct 92 04:59:00 -0400
From: [e-mail] (Richard Stallman)
To: [e-mail]
Cc: [e-mail]
Subject: action expressions

GNU C supports nested functions.  In 2.3, these should work in
Objective C.  This supplies most of the mechanism for creating action
objects if you want them.  In fact, I expect one can go the rest of
the way with macros, if you don't mind a different syntax:

#define make_action(name, body)
 ({ id __temp = <create an empty action object>;
    void name () { body }
    <store &name into __temp>;
    __temp; })

I've omitted the backslashes, and left stubs for where Objective C
constructs are needed since I don't know them.

This technique requires you to give a unique name each time you write
an action expression.  That could be avoided with an improvement in
macro power.

Date: Thu, 15 Oct 92 05:02:24 -0400
From: [e-mail] (Richard Stallman)
To: [e-mail]
Cc: [e-mail]
Subject: exceptions

There seems to be a concensus that (1) exceptions should work by
exiting to the level where the handler was set up, and (2) running a
debugger is a completely independent matter from handling exceptions.

I've seen systems where there was an attempt to permit handlers to
look at the environment of the exception.  The problem is that the
handler can't do anything useful except a goto, unless it understands
what was going on at the place that got the exception.  And the whole
idea is that the handler shouldn't have to know that.  So an exception
facility that always does a goto is sufficient.

There are various plans to implement exception handling in GCC for the
sake of various languages, and it will surely get done sooner or later.
It will be easy to support in Objective C once the mechanism is present.

Date: Thu, 15 Oct 92 05:25:13 -0400
From: [e-mail] (Richard Stallman)
To: [e-mail]
Cc: [e-mail], [e-mail]
Subject: exceptions

I see my message was not clearly written and might have been
confusing.

There seems to be a consensus among most language designers that (1)
exceptions should work by exiting to the level where the handler was
set up, and (2) running a debugger is a completely independent matter
from handling exceptions.  Brad Cox seems to disagree--but I agree
with the general consensus, and that's how I plan to support
exceptions in GCC.

Date: Thu, 15 Oct 1992 08:15:44 -0500
To: [e-mail] (Richard Stallman)
From: [e-mail] (Brad Cox)
Subject: Re: exceptions
Cc: [e-mail], [e-mail]

>There seems to be a consensus among most language designers that (1)
>exceptions should work by exiting to the level where the handler was
>set up, and (2) running a debugger is a completely independent matter
>from handling exceptions.

The problem is most obvious when designing a 'default' exception handler;
i.e. a piece of code internal to the EH package that will automatically
catch exceptions globally if the programmer doesn't provide one of his own.

This global handler needs to run in the context of the exception to, for
example, print a backtrace showing the calls that led to the exception. The
same applies if somebody wants to override the exception to call a more
sophisticated process inspector ( a 'debugger'), perhaps capable of
examining and possibly even changing global shared state (such as the
stack) and maybe even resuming the execution. For example, imagine invoking
adb (or something better) from inside the handler.

I suspect that the concensus you speak of arose from the difficulty of
writing exception handlers as subroutines rather than from more considered
reasons. Action expressions are useful for many other things than exception
handling. But once they become available, it becomes practical to
reconsider solutions that weren't practical in their absence...like running
exception handlers as subroutines of the exception. That is, exception
handlers often need to access state  local to the scope of the calling
site. This is clumsy if the handler is a subroutine, but easy if the
handler is an action object because all variables are copied into the
action object and passed to its handler as ordinary variables.

I've advocated a departure from concensus on this matter on the grounds of
generality...potentially useful (albeit nontrivial to access) information;
i.e. stack frames, are unconditionally discarded in the concensus solution
that is retained by the action expression approach.

And oh yes, the concensus approach is a concensus only for compiled
stack-based languages like C, but not for non-stack-based languages like
Smalltalk and (although I'm less sure of this; any Lisp experts in the
crowd?) and Lisp.
Brad Cox; 703 691 3187 direct; 703 993 1142 reception;
[e-mail]

From: [e-mail] (Bill Burcham)
Date: Thu, 15 Oct 92 09:59:06 -0500
To: [e-mail] (Brad Cox)
Subject: Re: exceptions
Cc: [e-mail]

Yeah, and there are also the problems of breaking a thread and never
getting back from traditional exceptions.  Also, NeXT's approach uses
int's to identify exceptions -- but they have no object to manage the
allocation of these ints, so if I write some reusable class, I have to
pick an int range at random (and pray that no one else has used or
will ever use any of those same ints for their exceptions).

If we choose the consensus scheme (not that I am saying we _should_), then I believe it is important to provide an ExceptionBroker:

+ (BOOL)setCodeBase:(int)base availableExceptions:(int)howmany;
+ registerErrorReporter:(void (*)(NXHandler *errorState))proc
    forException:(STR)aName;
+ raise:(STR)exception data1:(void*)data1 data2:(void*)data2;
+ (BOOL)senderMayRaise:(STR)aName;
+ (int)exceptionNumber:(STR)aName;

#define EB_RegisterErrorReporter(proc, name) \
[ExceptionBroker registerErrorReporter:(proc) \
    forException:(aName)]

#define EB_RAISE(exception, d1, d2) \
[ExceptionBroker raise:(exception) data1:(d1) data2:(d2)]

This protocol allows a class which will raise certain exceptions (or
will produce instances which will produce certain exceptions) to check
in +initialize (or -init) whether or not those exception identifiers
are available to it.  If they are not, then the class or instance can
set an ivar such as `BOOL classIsBroker' which can prevent subsequent
instantiation, etc.

I will donate ExceptionBroker and helper class Exception to The
Project if anyone is interested.

Bill

Date: Thu, 15 Oct 92 15:00:35 -0400
From: [e-mail] (Richard Stallman)
To: [e-mail]
Cc: [e-mail], [e-mail]
Subject: exceptions

    I suspect that the concensus you speak of arose from the difficulty of
    writing exception handlers as subroutines

The difficulty you speak of is limited to C.  Most of the language
design I'm talking about is for other languages that support nested
functions, and could easily use that mechanism for running an
exception handler as a subroutine, if that were desired.  There used
to be such designs, as in PL/1, but they were not very useful and that
led to the conclusion that handlers should always go back to the
context where they were set up.

    This global handler needs to run in the context of the exception to, for
    example, print a backtrace showing the calls that led to the exception.

The second part of the consensus is that handlers should not be the
basis for debugging.

It is not very useful to print a backtrace.  The ways we deal with
crashes is by either stopping the process or by making a core dump.
Either way, you can get a backtrace or do various other useful
things.  The way to do these is not with a handler.  Instead, the
function that raises an exception will raise a signal when
appropriate.


I'm not going to support more than one exception handler mechanism in
GCC unless there is a clear need.  I want to support just one, for all
languages--to save work.  It may be that this mechanism can handle
handlers that run as subroutines as well as handlers that jump,
without much extra trouble.  But I would still design the handler
construct in a new language to do jumps, at this point.