one big unit or several smaller units? (was: "Re: Adding support for VxWorks target")

[Joel Brobecker <brobecker@adacore.com>]

[changed the subject for this sub-thread]

> I skimmed, and didn't see anything problematic.

Thanks for looking at them!

> Organizationwise, my personal preference would to be have fewer files
> that are "everything wtx", which in turns reduces the amount of
> headers / header files.  It also forestalls some grumbles about long
> filenames from DOS-land. :-)  But don't feel compelled to glom them
> together, if you prefer multiple files.

It's interesting that you'd say that; I am the total opposite. I don't
mind the extra headers - I think it makes the public/private API clearer.

In fact, funny story: Jerome and I were in NY discussing some recent
breakage in the debugger, and we both said that, while GDB's code has
improved tremendously, some areas are still pretty horrible. After
being asked for examples, I mentioned a few areas, but after a couple
of minutes, Jerome said that one of the worst areas has got to be...
ada-lang! And, oh yeah, it has to be one of the worst... ;-)

And that brings me back to the original thought. I have always wanted
to spend the time to break ada-lang into smaller modules. For instance,
one module to deal with the GNAT encoding and provide a layer that
insulates the rest of the code from it. Another module to handle
expression evaluation. etc.  Given that ada-lang is over 10,000 lines
of code (actually 13,000 lines), I think it's pretty much the only way
I can have a full understanding of that code...

-- 
Joel