[fpc-pascal] Known issue with 1.9.4 [2004/05/30] for i386 on Linux?

[Alan Mead]

Alan Mead <cubrewer at yahoo.com> wrote:

> I don't know if this is really a bug or whether it's known... I
> have
> this line of code:
> 
>   while ( (i<=Length(fmt)) and (NOT fmt[i] IN ['0'..'9']) ) do
> inc(i);
> 
> It compiles fine but I get a RTE:
> 
> An unhandled exception occurred at 0x0807B180 :
> EVariantError : Invalid variant operation
>   $0807B180
>   $BFFFF980
> 
> Funny thing is, I enabled line numbers with '-gl'... And the error
> message makes no sense to me.. After staring at that line of code
> for
> a while, I put parens around the "fmt[i] IN ['0'..'9']" and all was
> well:
> 
>   while ( (i<=Length(fmt)) and (NOT (fmt[i] IN ['0'..'9'])) ) do
> inc(i);
> 
> So, that seems to be a bug to me... Shall I report it or is it
> known
> (or am I wrong that it's a bug)?

Thanks for the comments.  Perhaps I was unclear. There are three
reasons I am inclined to consider this a bug. 

First, it seems improper for any syntax to compile (i.e., to fail to
raise a syntax error) and then not execute. Ok, "a := 1/0;" might
compile but it seems like some type checking got lost here.  Maybe
there is am implicit conversion going on... At the least, I'd like
the compiler to flag this line with a warning.

Second, I had to track this down by inserting a bunch of lines like
"writeln('Passed 33');".  The code was compiled with -gl so I should
see a line number but none is shown.  This seems like a bug.  

Third, the RTE, "Invalid variant operation" seems inappropriate and
confusing.  Again, maybe there's an implicit conversion... (but is it
appropriate?)

-Alan