In article ,
Ruvim wrote:
>On 2025-05-06 14:12, dxf wrote:
>> Irrespective of the merits such a change to the spec for THROW would
>> be not be practical given most systems appear to use ANS' QUIT-based
>> exception handler. But for new implementers or the curious here is
>> how I organised mine and which is readily changeable.
>>
>>
>> \ return to OS with exit code
>> : RETURN ( code -- ) ... ;
>>
>> \ perform ANS QUIT
>> : (quit) ( -- ) r0 @ rp! reset normal /interpret
>> begin cr (refill) drop interpret state? 0= if (vstat)
>> @execute then again ;
>>
>> \ exit to OS or back to forth
>> : ?return ( code -- ) turnkey? if return then drop (quit) ;
>
>Returning to OS can be useful not only for a turnkey program, but also
>for batch mode, in which the Forth system must return to the OS on any
>uncaught error.
>
>A possible use-case in Linux shell:
>
> ./generate-forth-program params | forth --batch-mode
>
>since you probably prefer not to interpret the following lines if there
>an error occurs when defining a word in some line.
>
>
>>
>> \ clear data stacks
>> : (abort) ( i*x -- ) s0 @ sp! fs0 @ fsp ! 1 ?return ;
>>
>> \ part of THROW
>> : error ( n -- )
>> -1 of (abort) then
>> -2 of boot cell+ @ 0= if .error then
>> space errmsg 2@ type (abort) then
>> ." THROW #" @base decimal swap . !base (abort) ;
>>
>> : QUIT ( -- ) 0 ?return ; \ QUIT not trapped
>>
>> \ QUIT is not trapped in DX-Forth but may be made so by
>> \ adding -56 of 1 ?return then to 'error' and defining
>> \ : QUIT -56 throw ;
>
>
>I use `quit` only for testing and debugging. And I expect to get into
>the Forth text interpreter (Forth shell) exactly in the current program
>state (as far as possible).
>
>If `quit` simply calls `-56 throw`, the program state will probably
>change (due to actions after `catch` in the program, including restoring
>the data stack depth in `throw`) when you get into the Forth shell.
>Also, there is a chance that you will not get into the Forth shell at
>all if the program does not re-throw the error in some places.
>
>So I would not recommend the suggested deviation in the `quit` behavior
>even for new implementers, since this brakes the well-known expectation
>from `quit`. A better way is to introduce another word with desired
>behavior deviation.
I introduced `break that does deviate from `QUIT only in not cleaning
the return stack. So you can debug from such a nested "shell" with
impunity. A soon as you type ^D into this shell, you can continue
as if nothing happened. Or you can amend the stack, maybe even the
return stack, is you know what you're doing.
Of course `break must save and restore the current input stream.
I have words for that. It is straightforward. Others have
proposed INTERPRET-WHILE-READING or some such words.
Maybe these words could define a `break, but not as easily.
: break
SAVE
BEGIN '(ACCEPT) CATCH DUP E_BROKEN_PIPE <> WHILE
?ERRUR SET-SRC INTERPRET REPEAT
DROP RESTORE
: QUIT
POSTPONE [
BEGIN
R0 @ RSP! \ this part is missing from break.
'(ACCEPT) CATCH DUP E_BROKEN_PIPE = IF
BYE THEN
?ERRUR 1+ SET-SRC INTERPRET OK AGAIN
'(ACCEPT) CATCH
is approximately
REFILL
>
>
>
>>
>> : ?ABORT ( flag c-addr u -- )
>> rot if errmsg 2! -2 throw then 2drop ;
>>
>> : (abort") ( flag -- ) r> count 2dup + >r ?abort ;
>>
>> : ABORT" state @ if postpone (abort") ," end
>> postpone s" ?abort ; immediate
>>
>> : ABORT -1 throw ;
>>
>>
I have banned ABORT" c.s to the source library.
>--
>Ruvim
>
Groetjes Albert
--
Temu exploits Christians: (Disclaimer, only 10 apostles)
Last Supper Acrylic Suncatcher - 15Cm Round Stained Glass- Style Wall
Art For Home, Office And Garden Decor - Perfect For Windows, Bars,
And Gifts For Friends Family And Colleagues.
--- SoupGate-DOS v1.05
* Origin: you cannot sedate... all the things you hate (1:229/2)
|
