[M3devel] turning off checking?

[Elmar Stellnberger]

It would be a particularely bad idea to turn off checking as a whole for 
most use cases; i.e. f.i. just because there is one code snipped which 
does not work with these checks. It should be possible to switch of 
range checks i.e. integer underflow and overflow checks for a paticular 
code block; there could be a different switch for array index range 
checks and so on. Concerning the automaton simulator which I had started 
to port to Modula-3 last year there was one such snippet which could 
only be made working by an explicit type conversion to longint, applying 
a logical and and then converting back to integer - and that for several 
times. The code has become almost unreadable by the many type 
conversions which are basically unnecessary and degrading performance. 
Consequently I would really appreciate such switches on a per-code block 
level.

Elmar

Am 24.07.14 22:06, schrieb Rodney M. Bates:
>
>
> On 07/24/2014 05:08 AM, mika at async.caltech.edu wrote:
>> I added M3_OPTIONS=["NoChecks"] to my cm3.cfg:
>>
>> (62)truffles:~/t/btc/hsim/gcrit/src>cm3 -x
>> --- building in ../AMD64_LINUX ---
>>
>> new source -> compiling CardCritAttrTbl.i3
>> m3c: unknown option, "NoChecks"
>>
>
> Some grepping and reading suggest this message comes from 
> m3front/src/misc/Host.m3:158,
> procedure ProcessArg.  This would accept "-NoChecks", note the dash.
>
> Some more poking around reveals it gets these options from Quake 
> variables
> M3_OPTIONS and M3_FRONT_FLAGS.
>
> Despite the name "ProcessArgs" and the use of dashes on everything, these
> don't appear to be command line options, which are handled in entirely
> different paths.  Command line "-A" looks to get converted to 
> "-NoAsserts"
> and appended to M3_OPTIONS by generated Quake code, to be interpreted 
> later.
> There are a few other such conversions, but nothing else starting with 
> "-No..."
>
>> Fatal Error: failed compiling:
>>
>
> Snip -------
>