source: doc/theses/andrew_beach_MMath/features.tex @ a049412

arm-ehjacob/cs343-translationnew-ast-unique-expr
Last change on this file since a049412 was a6c45c6, checked in by Andrew Beach <ajbeach@…>, 7 months ago

Andrew MMath: Went over the features chapter again.

  • Property mode set to 100644
File size: 30.8 KB
Line 
1\chapter{Exception Features}
2
3This chapter covers the design and user interface of the \CFA
4exception-handling mechanism (EHM). % or exception system.
5
6We will begin with an overview of EHMs in general. It is not a strict
7definition of all EHMs nor an exaustive list of all possible features.
8However it does cover the most common structure and features found in them.
9
10% We should cover what is an exception handling mechanism and what is an
11% exception before this. Probably in the introduction. Some of this could
12% move there.
13\paragraph{Raise / Handle}
14An exception operation has two main parts: raise and handle.
15These terms are sometimes also known as throw and catch but this work uses
16throw/catch as a particular kind of raise/handle.
17These are the two parts that the user will write themselves and may
18be the only two pieces of the EHM that have any syntax in the language.
19
20\subparagraph{Raise}
21The raise is the starting point for exception handling. It marks the beginning
22of exception handling by raising an excepion, which passes it to
23the EHM.
24
25Some well known examples include the @throw@ statements of \Cpp and Java and
26the \codePy{raise} statement from Python. In real systems a raise may preform
27some other work (such as memory management) but for the purposes of this
28overview that can be ignored.
29
30\subparagraph{Handle}
31The purpose of most exception operations is to run some user code to handle
32that exception. This code is given, with some other information, in a handler.
33
34A handler has three common features: the previously mentioned user code, a
35region of code they cover and an exception label/condition that matches
36certain exceptions.
37Only raises inside the covered region and raising exceptions that match the
38label can be handled by a given handler.
39Different EHMs will have different rules to pick a handler
40if multipe handlers could be used such as ``best match" or ``first found".
41
42The @try@ statements of \Cpp, Java and Python are common examples. All three
43also show another common feature of handlers, they are grouped by the covered
44region.
45
46\paragraph{Propagation}
47After an exception is raised comes what is usually the biggest step for the
48EHM: finding and setting up the handler. The propogation from raise to
49handler can be broken up into three different tasks: searching for a handler,
50matching against the handler and installing the handler.
51
52\subparagraph{Searching}
53The EHM begins by searching for handlers that might be used to handle
54the exception. Searching is usually independent of the exception that was
55thrown as it looks for handlers that have the raise site in their covered
56region.
57This includes handlers in the current function, as well as any in callers
58on the stack that have the function call in their covered region.
59
60\subparagraph{Matching}
61Each handler found has to be matched with the raised exception. The exception
62label defines a condition that be use used with exception and decides if
63there is a match or not.
64
65In languages where the first match is used this step is intertwined with
66searching, a match check is preformed immediately after the search finds
67a possible handler.
68
69\subparagraph{Installing}
70After a handler is chosen it must be made ready to run.
71The implementation can vary widely to fit with the rest of the
72design of the EHM. The installation step might be trivial or it could be
73the most expensive step in handling an exception. The latter tends to be the
74case when stack unwinding is involved.
75
76If a matching handler is not guarantied to be found the EHM will need a
77different course of action here in the cases where no handler matches.
78This is only required with unchecked exceptions as checked exceptions
79(such as in Java) can make than guaranty.
80This different action can also be installing a handler but it is usually an
81implicat and much more general one.
82
83\subparagraph{Hierarchy}
84A common way to organize exceptions is in a hierarchical structure.
85This is especially true in object-orientated languages where the
86exception hierarchy is a natural extension of the object hierarchy.
87
88Consider the following hierarchy of exceptions:
89\begin{center}
90\input{exception-hierarchy}
91\end{center}
92
93A handler labelled with any given exception can handle exceptions of that
94type or any child type of that exception. The root of the exception hierarchy
95(here \codeC{exception}) acts as a catch-all, leaf types catch single types
96and the exceptions in the middle can be used to catch different groups of
97related exceptions.
98
99This system has some notable advantages, such as multiple levels of grouping,
100the ability for libraries to add new exception types and the isolation
101between different sub-hierarchies.
102This design is used in \CFA even though it is not a object-orientated
103language; so different tools are used to create the hierarchy.
104
105% Could I cite the rational for the Python IO exception rework?
106
107\paragraph{Completion}
108After the handler has finished the entire exception operation has to complete
109and continue executing somewhere else. This step is usually simple,
110both logically and in its implementation, as the installation of the handler
111is usually set up to do most of the work.
112
113The EHM can return control to many different places,
114the most common are after the handler definition and after the raise.
115
116\paragraph{Communication}
117For effective exception handling, additional information is usually passed
118from the raise to the handler.
119So far only communication of the exceptions' identity has been covered.
120A common method is putting fields into the exception instance and giving the
121handler access to them.
122
123\section{Virtuals}
124Virtual types and casts are not part of \CFA's EHM nor are they required for
125any EHM.
126However the \CFA uses a hierarchy built with the virtual system as the basis
127for exceptions and exception matching.
128
129The virtual system would have ideally been part of \CFA before the work
130on exception handling began, but unfortunately it was not.
131Because of this only the features and framework needed for the EHM were
132designed and implemented. Other features were considered to ensure that
133the structure could accomidate other desirable features but they were not
134implemented.
135The rest of this section will only discuss the finalized portion of the
136virtual system.
137
138The virtual system supports multiple ``trees" of types. Each tree is
139a simple hierarchy with a single root type. Each type in a tree has exactly
140one parent -- except for the root type which has zero parents -- and any
141number of children.
142Any type that belongs to any of these trees is called a virtual type.
143
144% A type's ancestors are its parent and its parent's ancestors.
145% The root type has no ancestors.
146% A type's decendents are its children and its children's decendents.
147
148Every virtual type also has a list of virtual members. Children inherit
149their parent's list of virtual members but may add new members to it.
150It is important to note that these are virtual members, not virtual methods
151of object-orientated programming, and can be of any type.
152However, since \CFA has function pointers and they are allowed, virtual
153members can be used to mimic virtual methods.
154
155Each virtual type has a unique id.
156This unique id and all the virtual members are combined
157into a virtual table type. Each virtual type has a pointer to a virtual table
158as a hidden field.
159
160Up until this point the virtual system is similar to ones found in
161object-orientated languages but this where \CFA diverges. Objects encapsulate a
162single set of behaviours in each type, universally across the entire program,
163and indeed all programs that use that type definition. In this sense the
164types are ``closed" and cannot be altered.
165
166In \CFA types do not encapsulate any behaviour. Traits are local and
167types can begin to statify a trait, stop satifying a trait or satify the same
168trait in a different way at any lexical location in the program.
169In this sense they are ``open" as they can change at any time. This means it
170is implossible to pick a single set of functions that repersent the type's
171implementation across the program.
172
173\CFA side-steps this issue by not having a single virtual table for each
174type. A user can define virtual tables which are filled in at their
175declaration and given a name. Anywhere that name is visible, even if it was
176defined locally inside a function (although that means it will not have a
177static lifetime), it can be used.
178Specifically, a virtual type is ``bound" to a virtual table which
179sets the virtual members for that object. The virtual members can be accessed
180through the object.
181
182While much of the virtual infrastructure is created, it is currently only used
183internally for exception handling. The only user-level feature is the virtual
184cast, which is the same as the \Cpp \codeCpp{dynamic_cast}.
185\label{p:VirtualCast}
186\begin{cfa}
187(virtual TYPE)EXPRESSION
188\end{cfa}
189Note, the syntax and semantics matches a C-cast, rather than the function-like
190\Cpp syntax for special casts. Both the type of @EXPRESSION@ and @TYPE@ must be
191a pointer to a virtual type.
192The cast dynamically checks if the @EXPRESSION@ type is the same or a sub-type
193of @TYPE@, and if true, returns a pointer to the
194@EXPRESSION@ object, otherwise it returns @0p@ (null pointer).
195
196\section{Exception}
197% Leaving until later, hopefully it can talk about actual syntax instead
198% of my many strange macros. Syntax aside I will also have to talk about the
199% features all exceptions support.
200
201Exceptions are defined by the trait system; there are a series of traits, and
202if a type satisfies them, then it can be used as an exception. The following
203is the base trait all exceptions need to match.
204\begin{cfa}
205trait is_exception(exceptT &, virtualT &) {
206        // Numerous imaginary assertions.
207};
208\end{cfa}
209The trait is defined over two types, the exception type and the virtual table
210type. Each exception type should have but a single virtual table type.
211Now there are no actual assertions in this trait because the trait system
212actually can't express them (adding such assertions would be part of
213completing the virtual system). The imaginary assertions would probably come
214from a trait defined by the virtual system, and state that the exception type
215is a virtual type, is a decendent of @exception_t@ (the base exception type)
216and note its virtual table type.
217
218% I did have a note about how it is the programmer's responsibility to make
219% sure the function is implemented correctly. But this is true of every
220% similar system I know of (except Agda's I guess) so I took it out.
221
222There are two more traits for exceptions defined as follows:
223\begin{cfa}
224trait is_termination_exception(
225                exceptT &, virtualT & | is_exception(exceptT, virtualT)) {
226        void defaultTerminationHandler(exceptT &);
227};
228
229trait is_resumption_exception(
230                exceptT &, virtualT & | is_exception(exceptT, virtualT)) {
231        void defaultResumptionHandler(exceptT &);
232};
233\end{cfa}
234Both traits ensure a pair of types are an exception type and its virtual table
235and defines one of the two default handlers. The default handlers are used
236as fallbacks and are discussed in detail in \vref{s:ExceptionHandling}.
237
238However, all three of these traits can be tricky to use directly.
239While there is a bit of repetition required,
240the largest issue is that the virtual table type is mangled and not in a user
241facing way. So these three macros are provided to wrap these traits to
242simplify referring to the names:
243@IS_EXCEPTION@, @IS_TERMINATION_EXCEPTION@ and @IS_RESUMPTION_EXCEPTION@.
244
245All three take one or two arguments. The first argument is the name of the
246exception type. The macro passes its unmangled and mangled form to the trait.
247The second (optional) argument is a parenthesized list of polymorphic
248arguments. This argument is only used with polymorphic exceptions and the
249list is be passed to both types.
250In the current set-up, the two types always have the same polymorphic
251arguments so these macros can be used without losing flexibility.
252
253For example consider a function that is polymorphic over types that have a
254defined arithmetic exception:
255\begin{cfa}
256forall(Num | IS_EXCEPTION(Arithmetic, (Num)))
257void some_math_function(Num & left, Num & right);
258\end{cfa}
259
260\section{Exception Handling}
261\label{s:ExceptionHandling}
262\CFA provides two kinds of exception handling: termination and resumption.
263These twin operations are the core of \CFA's exception handling mechanism.
264This section will cover the general patterns shared by the two operations and
265then go on to cover the details each individual operation.
266
267Both operations follow the same set of steps.
268Both start with the user preforming a raise on an exception.
269Then the exception propogates up the stack.
270If a handler is found the exception is caught and the handler is run.
271After that control returns to normal execution.
272If the search fails a default handler is run and then control
273returns to normal execution after the raise.
274
275This general description covers what the two kinds have in common.
276Differences include how propogation is preformed, where exception continues
277after an exception is caught and handled and which default handler is run.
278
279\subsection{Termination}
280\label{s:Termination}
281Termination handling is the familiar kind and used in most programming
282languages with exception handling.
283It is dynamic, non-local goto. If the raised exception is matched and
284handled the stack is unwound and control will (usually) continue the function
285on the call stack that defined the handler.
286Termination is commonly used when an error has occurred and recovery is
287impossible locally.
288
289% (usually) Control can continue in the current function but then a different
290% control flow construct should be used.
291
292A termination raise is started with the @throw@ statement:
293\begin{cfa}
294throw EXPRESSION;
295\end{cfa}
296The expression must return a reference to a termination exception, where the
297termination exception is any type that satisfies the trait
298@is_termination_exception@ at the call site.
299Through \CFA's trait system the trait functions are implicity passed into the
300throw code and the EHM.
301A new @defaultTerminationHandler@ can be defined in any scope to
302change the throw's behavior (see below).
303
304The throw will copy the provided exception into managed memory to ensure
305the exception is not destroyed if the stack is unwound.
306It is the user's responsibility to ensure the original exception is cleaned
307up wheither the stack is unwound or not. Allocating it on the stack is
308usually sufficient.
309
310Then propogation starts with the search. \CFA uses a ``first match" rule so
311matching is preformed with the copied exception as the search continues.
312It starts from the throwing function and proceeds to the base of the stack,
313from callee to caller.
314At each stack frame, a check is made for resumption handlers defined by the
315@catch@ clauses of a @try@ statement.
316\begin{cfa}
317try {
318        GUARDED_BLOCK
319} catch (EXCEPTION_TYPE$\(_1\)$ * [NAME$\(_1\)$]) {
320        HANDLER_BLOCK$\(_1\)$
321} catch (EXCEPTION_TYPE$\(_2\)$ * [NAME$\(_2\)$]) {
322        HANDLER_BLOCK$\(_2\)$
323}
324\end{cfa}
325When viewed on its own, a try statement will simply execute the statements
326in @GUARDED_BLOCK@ and when those are finished the try statement finishes.
327
328However, while the guarded statements are being executed, including any
329invoked functions, all the handlers in the statement are now on the search
330path. If a termination exception is thrown and not handled further up the
331stack they will be matched against the exception.
332
333Exception matching checks the handler in each catch clause in the order
334they appear, top to bottom. If the representation of the thrown exception type
335is the same or a descendant of @EXCEPTION_TYPE@$_i$ then @NAME@$_i$
336(if provided) is
337bound to a pointer to the exception and the statements in @HANDLER_BLOCK@$_i$
338are executed. If control reaches the end of the handler, the exception is
339freed and control continues after the try statement.
340
341If no termination handler is found during the search then the default handler
342(@defaultTerminationHandler@) is run.
343Through \CFA's trait system the best match at the throw sight will be used.
344This function is run and is passed the copied exception. After the default
345handler is run control continues after the throw statement.
346
347There is a global @defaultTerminationHandler@ that is polymorphic over all
348exception types. Since it is so general a more specific handler can be
349defined and will be used for those types, effectively overriding the handler
350for particular exception type.
351The global default termination handler performs a cancellation
352(see \vref{s:Cancellation}) on the current stack with the copied exception.
353
354\subsection{Resumption}
355\label{s:Resumption}
356
357Resumption exception handling is less common than termination but is
358just as old~\cite{Goodenough75} and is simpler in many ways.
359It is a dynamic, non-local function call. If the raised exception is
360matched a closure will be taken from up the stack and executed,
361after which the raising function will continue executing.
362These are most often used when an error occurred and if the error is repaired
363then the function can continue.
364
365A resumption raise is started with the @throwResume@ statement:
366\begin{cfa}
367throwResume EXPRESSION;
368\end{cfa}
369It works much the same way as the termination throw.
370The expression must return a reference to a resumption exception,
371where the resumption exception is any type that satisfies the trait
372@is_resumption_exception@ at the call site.
373The assertions from this trait are available to
374the exception system while handling the exception.
375
376At run-time, no exception copy is made.
377As the stack is not unwound the exception and
378any values on the stack will remain in scope while the resumption is handled.
379
380The EHM then begins propogation. The search starts from the raise in the
381resuming function and proceeds to the base of the stack, from callee to caller.
382At each stack frame, a check is made for resumption handlers defined by the
383@catchResume@ clauses of a @try@ statement.
384\begin{cfa}
385try {
386        GUARDED_BLOCK
387} catchResume (EXCEPTION_TYPE$\(_1\)$ * [NAME$\(_1\)$]) {
388        HANDLER_BLOCK$\(_1\)$
389} catchResume (EXCEPTION_TYPE$\(_2\)$ * [NAME$\(_2\)$]) {
390        HANDLER_BLOCK$\(_2\)$
391}
392\end{cfa}
393% I wonder if there would be some good central place for this.
394Note that termination handlers and resumption handlers may be used together
395in a single try statement, intermixing @catch@ and @catchResume@ freely.
396Each type of handler will only interact with exceptions from the matching
397type of raise.
398When a try statement is executed it simply executes the statements in the
399@GUARDED_BLOCK@ and then finishes.
400
401However, while the guarded statements are being executed, including any
402invoked functions, all the handlers in the statement are now on the search
403path. If a resumption exception is reported and not handled further up the
404stack they will be matched against the exception.
405
406Exception matching checks the handler in each catch clause in the order
407they appear, top to bottom. If the representation of the thrown exception type
408is the same or a descendant of @EXCEPTION_TYPE@$_i$ then @NAME@$_i$
409(if provided) is bound to a pointer to the exception and the statements in
410@HANDLER_BLOCK@$_i$ are executed.
411If control reaches the end of the handler, execution continues after the
412the raise statement that raised the handled exception.
413
414Like termination, if no resumption handler is found, the default handler
415visible at the throw statement is called. It will use the best match at the
416call sight according to \CFA's overloading rules. The default handler is
417passed the exception given to the throw. When the default handler finishes
418execution continues after the raise statement.
419
420There is a global @defaultResumptionHandler@ is polymorphic over all
421termination exceptions and preforms a termination throw on the exception.
422The @defaultTerminationHandler@ for that raise is matched at the original
423raise statement (the resumption @throwResume@) and it can be customized by
424introducing a new or better match as well.
425
426\subsubsection{Resumption Marking}
427\label{s:ResumptionMarking}
428A key difference between resumption and termination is that resumption does
429not unwind the stack. A side effect that is that when a handler is matched
430and run it's try block (the guarded statements) and every try statement
431searched before it are still on the stack. This can lead to the recursive
432resumption problem.
433
434The recursive resumption problem is any situation where a resumption handler
435ends up being called while it is running.
436Consider a trivial case:
437\begin{cfa}
438try {
439        throwResume (E &){};
440} catchResume(E *) {
441        throwResume (E &){};
442}
443\end{cfa}
444When this code is executed the guarded @throwResume@ will throw, start a
445search and match the handler in the @catchResume@ clause. This will be
446call and placed on the stack on top of the try-block. The second throw then
447throws and will search the same try block and put call another instance of the
448same handler leading to an infinite loop.
449
450This situation is trivial and easy to avoid, but much more complex cycles
451can form with multiple handlers and different exception types.
452
453To prevent all of these cases we mark try statements on the stack.
454A try statement is marked when a match check is preformed with it and an
455exception. The statement will be unmarked when the handling of that exception
456is completed or the search completes without finding a handler.
457While a try statement is marked its handlers are never matched, effectify
458skipping over it to the next try statement.
459
460\begin{center}
461\input{stack-marking}
462\end{center}
463
464These rules mirror what happens with termination.
465When a termination throw happens in a handler the search will not look at
466any handlers from the original throw to the original catch because that
467part of the stack has been unwound.
468A resumption raise in the same situation wants to search the entire stack,
469but it will not try to match the exception with try statements in the section
470that would have been unwound as they are marked.
471
472The symmetry between resumption termination is why this pattern was picked.
473Other patterns, such as marking just the handlers that caught, also work but
474lack the symmetry means there are more rules to remember.
475
476\section{Conditional Catch}
477Both termination and resumption handler clauses can be given an additional
478condition to further control which exceptions they handle:
479\begin{cfa}
480catch (EXCEPTION_TYPE * [NAME] ; CONDITION)
481\end{cfa}
482First, the same semantics is used to match the exception type. Second, if the
483exception matches, @CONDITION@ is executed. The condition expression may
484reference all names in scope at the beginning of the try block and @NAME@
485introduced in the handler clause. If the condition is true, then the handler
486matches. Otherwise, the exception search continues as if the exception type
487did not match.
488
489The condition matching allows finer matching by allowing the match to check
490more kinds of information than just the exception type.
491\begin{cfa}
492try {
493        handle1 = open( f1, ... );
494        handle2 = open( f2, ... );
495        handle3 = open( f3, ... );
496        ...
497} catch( IOFailure * f ; fd( f ) == f1 ) {
498        // Only handle IO failure for f1.
499} catch( IOFailure * f ; fd( f ) == f3 ) {
500        // Only handle IO failure for f3.
501}
502// Can't handle a failure relating to f2 here.
503\end{cfa}
504In this example the file that experianced the IO error is used to decide
505which handler should be run, if any at all.
506
507\begin{comment}
508% I know I actually haven't got rid of them yet, but I'm going to try
509% to write it as if I had and see if that makes sense:
510\section{Reraising}
511\label{s:Reraising}
512Within the handler block or functions called from the handler block, it is
513possible to reraise the most recently caught exception with @throw@ or
514@throwResume@, respectively.
515\begin{cfa}
516try {
517        ...
518} catch( ... ) {
519        ... throw;
520} catchResume( ... ) {
521        ... throwResume;
522}
523\end{cfa}
524The only difference between a raise and a reraise is that reraise does not
525create a new exception; instead it continues using the current exception, \ie
526no allocation and copy. However the default handler is still set to the one
527visible at the raise point, and hence, for termination could refer to data that
528is part of an unwound stack frame. To prevent this problem, a new default
529handler is generated that does a program-level abort.
530\end{comment}
531
532\subsection{Comparison with Reraising}
533A more popular way to allow handlers to match in more detail is to reraise
534the exception after it has been caught if it could not be handled here.
535On the surface these two features seem interchangable.
536
537If we used @throw;@ to start a termination reraise then these two statements
538would have the same behaviour:
539\begin{cfa}
540try {
541    do_work_may_throw();
542} catch(exception_t * exc ; can_handle(exc)) {
543    handle(exc);
544}
545\end{cfa}
546
547\begin{cfa}
548try {
549    do_work_may_throw();
550} catch(exception_t * exc) {
551    if (can_handle(exc)) {
552        handle(exc);
553    } else {
554        throw;
555    }
556}
557\end{cfa}
558If there are further handlers after this handler only the first version will
559check them. If multiple handlers on a single try block that could handle the
560same exception the translations get more complex but they are equivilantly
561powerful.
562
563Until stack unwinding comes into the picture. In termination handling, a
564conditional catch happens before the stack is unwound, but a reraise happens
565afterwards. Normally this might only cause you to loose some debug
566information you could get from a stack trace (and that can be side stepped
567entirely by collecting information during the unwind). But for \CFA there is
568another issue, if the exception isn't handled the default handler should be
569run at the site of the original raise.
570
571There are two problems with this: the site of the original raise doesn't
572exist anymore and the default handler might not exist anymore. The site will
573always be removed as part of the unwinding, often with the entirety of the
574function it was in. The default handler could be a stack allocated nested
575function removed during the unwind.
576
577This means actually trying to pretend the catch didn't happening, continuing
578the original raise instead of starting a new one, is infeasible.
579That is the expected behaviour for most languages and we can't replicate
580that behaviour.
581
582\section{Finally Clauses}
583\label{s:FinallyClauses}
584Finally clauses are used to preform unconditional clean-up when leaving a
585scope and are placed at the end of a try statement after any handler clauses:
586\begin{cfa}
587try {
588        GUARDED_BLOCK
589} ... // any number or kind of handler clauses
590... finally {
591        FINALLY_BLOCK
592}
593\end{cfa}
594The @FINALLY_BLOCK@ is executed when the try statement is removed from the
595stack, including when the @GUARDED_BLOCK@ finishes, any termination handler
596finishes or during an unwind.
597The only time the block is not executed is if the program is exited before
598the stack is unwound.
599
600Execution of the finally block should always finish, meaning control runs off
601the end of the block. This requirement ensures control always continues as if
602the finally clause is not present, \ie finally is for cleanup not changing
603control flow.
604Because of this requirement, local control flow out of the finally block
605is forbidden. The compiler precludes any @break@, @continue@, @fallthru@ or
606@return@ that causes control to leave the finally block. Other ways to leave
607the finally block, such as a long jump or termination are much harder to check,
608and at best requiring additional run-time overhead, and so are only
609discouraged.
610
611Not all languages with unwinding have finally clauses. Notably \Cpp does
612without it as descructors serve a similar role. Although destructors and
613finally clauses can be used in many of the same areas they have their own
614use cases like top-level functions and lambda functions with closures.
615Destructors take a bit more work to set up but are much easier to reuse while
616finally clauses are good for one-off uses and
617can easily include local information.
618
619\section{Cancellation}
620\label{s:Cancellation}
621Cancellation is a stack-level abort, which can be thought of as as an
622uncatchable termination. It unwinds the entire current stack, and if
623possible forwards the cancellation exception to a different stack.
624
625Cancellation is not an exception operation like termination or resumption.
626There is no special statement for starting a cancellation; instead the standard
627library function @cancel_stack@ is called passing an exception. Unlike a
628raise, this exception is not used in matching only to pass information about
629the cause of the cancellation.
630(This also means matching cannot fail so there is no default handler.)
631
632After @cancel_stack@ is called the exception is copied into the EHM's memory
633and the current stack is
634unwound. After that it depends one which stack is being cancelled.
635
636\paragraph{Main Stack}
637The main stack is the one used by the program main at the start of execution,
638and is the only stack in a sequential program.
639After the main stack is unwound there is a program-level abort.
640
641There are two reasons for this. The first is that it obviously had to do this
642in a sequential program as there is nothing else to notify and the simplicity
643of keeping the same behaviour in sequential and concurrent programs is good.
644Also, even in concurrent programs there is no stack that an innate connection
645to, so it would have be explicitly managed.
646
647\paragraph{Thread Stack}
648A thread stack is created for a \CFA @thread@ object or object that satisfies
649the @is_thread@ trait.
650After a thread stack is unwound there exception is stored until another
651thread attempts to join with it. Then the exception @ThreadCancelled@,
652which stores a reference to the thread and to the exception passed to the
653cancellation, is reported from the join.
654There is one difference between an explicit join (with the @join@ function)
655and an implicit join (from a destructor call). The explicit join takes the
656default handler (@defaultResumptionHandler@) from its calling context while
657the implicit join provides its own which does a program abort if the
658@ThreadCancelled@ exception cannot be handled.
659
660Communication is done at join because a thread only has to have to points of
661communication with other threads: start and join.
662Since a thread must be running to perform a cancellation (and cannot be
663cancelled from another stack), the cancellation must be after start and
664before the join. So join is the one that we will use.
665
666% TODO: Find somewhere to discuss unwind collisions.
667The difference between the explicit and implicit join is for safety and
668debugging. It helps prevent unwinding collisions by avoiding throwing from
669a destructor and prevents cascading the error across multiple threads if
670the user is not equipped to deal with it.
671Also you can always add an explicit join if that is the desired behaviour.
672
673\paragraph{Coroutine Stack}
674A coroutine stack is created for a @coroutine@ object or object that
675satisfies the @is_coroutine@ trait.
676After a coroutine stack is unwound control returns to the resume function
677that most recently resumed it. The resume statement reports a
678@CoroutineCancelled@ exception, which contains a references to the cancelled
679coroutine and the exception used to cancel it.
680The resume function also takes the @defaultResumptionHandler@ from the
681caller's context and passes it to the internal report.
682
683A coroutine knows of two other coroutines, its starter and its last resumer.
684The starter has a much more distant connection while the last resumer just
685(in terms of coroutine state) called resume on this coroutine, so the message
686is passed to the latter.
Note: See TracBrowser for help on using the repository browser.