Subj : Re: Compiler and an interpreter
To   : comp.programming
From : Rob Thorpe
Date : Sun Jul 31 2005 02:01 pm

Gerry Quinn wrote:
> In article <42eb8672$0$14673$ed2619ec@ptn-nntp-reader02.plus.net>,
> usenet@jdh30.plus.com says...
> > Gerry Quinn wrote:
> > > Typically an interpreter will stop on the first error.  Compilers in my
> > > experience usually carry on until a fixed number of errors (e.g. 100)
> > > have been found, and then stop.
> >
> > Lots of compilers don't do that.
> >
> > > Usually in C++, if there are eight errors found, there really are about
> > > eight.
> >
> > C++ isn't worth using unless you use the STL, IMHO. Error messages from all
> > of the compilers that I have seen are completely unwieldy when using the
> > STL. For example, the following code contains an extra "const":
>
> STL is just a library.  I got on well enough with the MFC collection
> classes for a while, although I do prefer STL now.  I still use MFC
> CString.  The rest of the stuff is just functions you might need once
> in a while.
>
> [STL example - I'm not sure I would call it C++]
>
> > > When there are 100 errors, you left out a curly bracket
> > > somewhere above the first one, or something like that.
> >
> > What you are saying is true in C but not in C++.
>
> It's true in C++ in my experience - STL errors are not as common as
> simple syntax errors.  Perhaps if you are trying to emulate some kind
> of functional language with C++ you will get more.
>
> There is a freeware add-on called STLFilt that is supposed to parse
> these errors for common compilers - haven't tried it.  At the end of
> the day an uninformative error message still does most of its job as it
> tells you there is an error and approximately where.

The C++ compilers I've used in the past all gave extremely
incomprehensible messages.  GCC was particularly bad, I believe it's
much better now.

The best strategy I found was to ignore the error message entirely
except the line number, then go to that line and look at it to see
what's wrong.

Even if you understand the principle of the error message, which is
sometimes not hard, it takes a bit of thinking time to apply it in
every circumstance.  So, my attitude was I'd rather be thinking about
the code instead.

Like Jon, I also wrote a many STL errors when I did C++.  It depends on
how the language is used, if you only use string, vector and valarray
there not much to go wrong, but if you go further ...

.