[PATCH] interpreter.fs: allow evaluate to split words on CR as well as LF
Otherwise the Forth intepreter fails due to lack of buffer space when trying to execute large boot scripts on platforms that use CR instead of LF for line endings (particularly MacOS 9). Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk> --- openbios-devel/forth/bootstrap/interpreter.fs | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/openbios-devel/forth/bootstrap/interpreter.fs b/openbios-devel/forth/bootstrap/interpreter.fs index 5187058..74a109f 100644 --- a/openbios-devel/forth/bootstrap/interpreter.fs +++ b/openbios-devel/forth/bootstrap/interpreter.fs @@ -163,7 +163,7 @@ defer outer-interpreter : evaluate ( str len -- ?? ) 2dup + -rot over + over do - i c@ 0a = if + i c@ dup 0a = swap 0d = or if i over - (evaluate) i 1+ -- 1.7.10.4
On 18/06/15 23:36, Mark Cave-Ayland wrote:
Otherwise the Forth intepreter fails due to lack of buffer space when trying to execute large boot scripts on platforms that use CR instead of LF for line endings (particularly MacOS 9).
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk> --- openbios-devel/forth/bootstrap/interpreter.fs | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/openbios-devel/forth/bootstrap/interpreter.fs b/openbios-devel/forth/bootstrap/interpreter.fs index 5187058..74a109f 100644 --- a/openbios-devel/forth/bootstrap/interpreter.fs +++ b/openbios-devel/forth/bootstrap/interpreter.fs @@ -163,7 +163,7 @@ defer outer-interpreter : evaluate ( str len -- ?? ) 2dup + -rot over + over do - i c@ 0a = if + i c@ dup 0a = swap 0d = or if i over - (evaluate) i 1+
While this seems to be correct, further testing shows that yaboot (which has some embedded \r codes in its Forth strings) emits warnings on the console with this patch applied. Will dig a bit deeper to try and figure out what is going on here. ATB, Mark.
On 20/06/15 12:47, Mark Cave-Ayland wrote:
On 18/06/15 23:36, Mark Cave-Ayland wrote:
Otherwise the Forth intepreter fails due to lack of buffer space when trying to execute large boot scripts on platforms that use CR instead of LF for line endings (particularly MacOS 9).
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk> --- openbios-devel/forth/bootstrap/interpreter.fs | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/openbios-devel/forth/bootstrap/interpreter.fs b/openbios-devel/forth/bootstrap/interpreter.fs index 5187058..74a109f 100644 --- a/openbios-devel/forth/bootstrap/interpreter.fs +++ b/openbios-devel/forth/bootstrap/interpreter.fs @@ -163,7 +163,7 @@ defer outer-interpreter : evaluate ( str len -- ?? ) 2dup + -rot over + over do - i c@ 0a = if + i c@ dup 0a = swap 0d = or if i over - (evaluate) i 1+
While this seems to be correct, further testing shows that yaboot (which has some embedded \r codes in its Forth strings) emits warnings on the console with this patch applied. Will dig a bit deeper to try and figure out what is going on here.
So the real problem here appears to be that the OS 9 bootscript is using the rstack in order to store temporary variables. Since the Forth version of interpret is based in a loop which looks up each word in turn and then runs execute, the manipulation of the rstack with r> confuses the loop causing it to drop out early. Actually I'm surprised we got as far as we did, however the OpenBIOS version of the execute word has a trampoline and combined with the fact that the OS 9 script has a CR just before the use of r> I think we were just extremely lucky. I think the real solution here is to evaluate the Forth string in compile mode to an anonymous word in the dictionary and then execute it directly, but I'd welcome any other ideas. ATB, Mark.
On Sat, Jun 20, 2015 at 08:41:28PM +0100, Mark Cave-Ayland wrote:
So the real problem here appears to be that the OS 9 bootscript is using the rstack in order to store temporary variables. Since the Forth version of interpret is based in a loop which looks up each word in turn and then runs execute, the manipulation of the rstack with r> confuses the loop causing it to drop out early.
You should use a separate (software) stack for rstack things in interpret mode (this is what other OF implementations do).
I think the real solution here is to evaluate the Forth string in compile mode to an anonymous word in the dictionary and then execute it directly, but I'd welcome any other ideas.
That isn't going to work, not without a lot more hackery (the script creates many definitions by itself). Segher
On 21/06/15 00:07, Segher Boessenkool wrote:
On Sat, Jun 20, 2015 at 08:41:28PM +0100, Mark Cave-Ayland wrote:
So the real problem here appears to be that the OS 9 bootscript is using the rstack in order to store temporary variables. Since the Forth version of interpret is based in a loop which looks up each word in turn and then runs execute, the manipulation of the rstack with r> confuses the loop causing it to drop out early.
You should use a separate (software) stack for rstack things in interpret mode (this is what other OF implementations do).
Okay. Don't suppose you could point me to some examples on how to do this? ATB, Mark.
On Sun, Jun 21, 2015 at 09:08:00AM +0100, Mark Cave-Ayland wrote:
So the real problem here appears to be that the OS 9 bootscript is using the rstack in order to store temporary variables. Since the Forth version of interpret is based in a loop which looks up each word in turn and then runs execute, the manipulation of the rstack with r> confuses the loop causing it to drop out early.
You should use a separate (software) stack for rstack things in interpret mode (this is what other OF implementations do).
Okay. Don't suppose you could point me to some examples on how to do this?
Oh, sorry. You could look at forth/lib/pseudors.fth in the OFW code. Segher
On 21/06/15 09:28, Segher Boessenkool wrote:
On Sun, Jun 21, 2015 at 09:08:00AM +0100, Mark Cave-Ayland wrote:
So the real problem here appears to be that the OS 9 bootscript is using the rstack in order to store temporary variables. Since the Forth version of interpret is based in a loop which looks up each word in turn and then runs execute, the manipulation of the rstack with r> confuses the loop causing it to drop out early.
You should use a separate (software) stack for rstack things in interpret mode (this is what other OF implementations do).
Okay. Don't suppose you could point me to some examples on how to do this?
Oh, sorry. You could look at forth/lib/pseudors.fth in the OFW code.
That's interesting. It looks like the above implementation allows different compile behaviour compared to OpenBIOS which always executes the underlying C function, regardless of what mode we are in. ATB, Mark.
participants (2)
-
Mark Cave-Ayland -
Segher Boessenkool