Subj : Re: Supressing errors?
To   : Brian Genisio <BrianGenisio@yahoo.com>
From : Brendan Eich <brendan@meer.net>
Date : Fri Jan 30 2004 11:58 am

Brian Genisio wrote:

> Hi all,
>
> I was wondering what the best way to supress specific errors are?  I 
> can think of couple ways to do it, but none are extremely keen in my 
> book.
>
> Here is where I want to supress the error:
>
> When a non-numeric string is passed to JS_ValueToInt32, an error is 
> reported, and it returns false.  All I care about is that we return 
> false.  If I get false, I will use a default value, and move on.


Do you want JS_ValueToECMAInt32 instead?  JS_ValueToInt32 range-errors 
if the value doesn't convert to a number, or if it converts to a number 
that doesn't fit in a 32-bit signed int; then it rounds to nearest.  
JS_ValueToECMAInt32 takes non-finite (NaN, +/-Infinity) as 0, then 
computes fmod of the number with 2^32, then truncates the fmod result 
toward zero.

The practical differences are:

1.  fmod vs. range error;
2.  round to nearest vs. truncation toward zero.

> I would rather not print an error when I use the method in this way, 
> but I dont know a _good_ way of detecting that we dont care about the 
> error. 


Error suppression is done in the engine, and can be done using the API.  
See JS_SetErrorReporter, JS_SaveExceptionState, 
JS_RestoreExceptionState, and JS_DropExceptionState.

/be

.