Re: [Lxr-dev] [ lxr-Bugs-519945 ] definitions are also references

[Shree K. <sh...@pr...>]

Hi Malcolm,

> I think this is correct - the symbol is referenced at line 1, it just
> happens that this reference is a definition.  Remember, the lxr doesn't
> attempt to do full parsing of the source (deliberately at the moment),
> so references simply means that the string "TEST" occurs here and looks
> like it might be a symbol.
> 
Right, but when you say "reference", most people tend to think that the 
symbol is being "used" in some way at that line [maybe as an
assignment].

It's highly unlikely that anyone would write code like

void func() { } void test { func() }

> Umm, why would a symbol be being defined many times?  Most programming
> languages don't allow you to define a symbol multiple times.
> 

One for the fn prototype, one for the fn declaration [think of this over
2 or
3 different platforms]. I've seen the problem becoming acute,
particularly
with #defines.

And it is difficult to filter out the definitions from the references
list,
when there are too many "definition"s.

In a source tree, a symbol being defined many times is a distinct
possiblility 
[I have struggled with this problem before]. Not everything in a source
tree 
might be linked into one.

> Personally I'm against this change, but what do others think?

Here's one reason that I wanted this change:

I've run into situations where we'd like to know where the variable
is being initialized, changed, etc. and it is so frustrating to 
jump to a reference and find out that it is not really a "reference".

- Shree Kumar