Re: looking for a c++ -> c++ parser

Greg Lindahl wrote:


> ... the EDG parser source
> is available free for non-commercial use, and I heard (a while ago)
> about a project by Kathy Lindlan <klindlan@cs.uoregon.edu> to build
> some stuff with the EDG parser which allows tool builders to do
> interesting things to parsed C++ programs. Details are at:
>
> http://www.cs.uoregon.edu/~klindlan/
>
> I'm not sure if there is a tool which takes her IL and outputs C++,
> but that part doesn't look hard.


It is true that the EDG parser can be used for these purposes; a
couple of years ago I worked for a company that used it in just that
way (inserting instrumentation into the intermediate tree, emitting
C++ source, then compiling that).


Solving the "emit valid C++ from an intermediate IL" can be a good
deal harder than you seem to think. Several bright people with years
of experience fiddling with the EDG tools did get it to apparently
work; there was some back-and-forth with the guys at EDG, and it is
probably easier now than it was when the project began. It is not
100% possible to add arbitrary instrumentation to C++; in some cases,
"there's no there there" to insert your instrumentation into (this
depends somewhat upon whether you wish to retain compatibility with
uninstrumented code; if you relax that restriction, life gets a good
deal easier). Lack of a well-defined evaluation order for C++
expressions (as in, ONE legal evaluation order) makes life tough, as
does different back-end compilers' ideas about temporary lifetimes (it
helps when they get it right). Buggy, and varying (that should go
away soon now that there's a standard, right?) C++ compilers were a
substantial headache.


The result of this experience building C++ debugging tools was a
substantial enthusiasm for Java :-).


David Chase
NaturalBridge LLC
(former employee of CenterLine Software)

Return to the comp.compilers page.
Search the comp.compilers archives again.