Gabriel Dos Reis wrote:
Robert Dewar <[EMAIL PROTECTED]> writes:
| It's interestinng to note that in the Ada world, there is an ISO
| standard for plugins, which is compiler/vendor neutral (at least
| in theory, in practice there are some implementation dependencies).
| That's the ASIS interface (Ada Semantic Interface Specification).
So, is it that plugins are good for Ada (and I presume the GNU Ada
front end) but not for GCC?
Well remember that the starting point in Ada that makes the plugin
useful is the ISO standard which achieves several things:
a) uniformity across different vendor implementations. It is possible
to write a semantic analysis tool that will run with GNAT or with
other Ada compilers implementing ASIS.
b) full formal documentation of the interface
Note that the existence of b) means clearly that the interface
is not proprietary, and in the case of GNAT not associated with
the GPL in any way in my non-lawyer, but offically-court-expert
view. There is a big difference between standard interfaces
of this kind and cooked up interfaces between two programs that
are dependent on one another to form a complete product.
To me it is premature to implement any kind of plug in without
complete documentation that gets extensively reviewed in advance
of any implementation, so I would hesitate to use the Ada ASIS
approach as justification for cooking up some idiosyncratic
not-completely-well-documented interface for gcc.