Leo --
Thats going to be a problem for Jako.
Can't nest .sub, so I can't have a bracketing .sub for the whole script,
with my real subs' .subs inside. So, I have to choose (I think) from these
options in order to leverage IMCC:
* Bracket with one big .sub ... .end, and don't use .sub for my real
subs (yuck -- I want to use .param, .arg, etc for subs and calls)
* Use the anonymous blocks to bracket all non-sub linear code in the
script as I have done (yuck -- results in the problems I've reported)
* Change the Jako language to require all linear code to be either
before or after all the subs (yuck -- Perl doesn't force this, and I don't
want to, either)
* Change the Jako compiler to collect all linear code into a single
segment and emit it at the end, inside .sub ... .end, with a preamble
..sub; goto __COLLECTED_LINEAR_CODE; .end at the beginning of the .imc file
(yuck -- To date, the Jako compiler has gotten away with doing its job as
a single pass essentially in filter mode, and I'd like to retain that
feature)
Given these 4 choices, I choose #5: try to change the rules of the game.
:)
I suggest that imcc should allow code outside .sub ... .end
brackets, which should be good for Jako, Perl and other languages.
I don't know how popular that will be, but it seems reasonable to me...
Regards,
-- Gregor
Leopold Toetsch <[EMAIL PROTECTED]>
12/05/2002 11:34 AM
To: [EMAIL PROTECTED]
cc: [EMAIL PROTECTED]
Subject: Re: Subroutine IMC example with problem
[EMAIL PROTECTED] wrote:
>
> error:imcc:iANY file sub.imc line 36: op not found 'set_ic_ic'
2 "mistakes" here:
- branches between subs are not subject to fixup
- x is not declared in here:
> .sub __ANON_BLOCK__1
> set x, 42 # x = 42
A .sub is a unit which is parsed/compiled in one piece. As I already
wrote: for your code put one .sub at the begin and one .end at the end -
that's all.
To see, what the parser thinks run: imcc -y x.imc 2>&1 | less
leo
