grauzone wrote:
To us, this doesn't really matter. The important thing is that we can
build the compiler itself, and can debug it if it craps up (which
happens often, sorry Walter). For example, now we might be able to find
out on which piece of code exactly it segfaults when compiling. Oh, and
remember the libc issues. Now users affected by this problem can compile
their own binary.
Making the source available deals with the following:
1. people can be assured there's no hanky-panky going on in the code -
no spyware, trojans, back doors, phone homing, etc.
2. if there's a serious problem, users can apply a patch themselves
rather than wait around for me
3. if Digital Mars vanishes in a puff of greasy black smoke, users
depending on dmd are not left holding the bag
4. makes custom ports the user cares about but nobody else does possible
5. engineers are curious and like to peek under the hood <g>.
6. hopefully it will make finding and correcting problems faster as
people can build it with symbolic debug info, and submit patches
7. allow for two way feedback on it, which should improve the quality
What it doesn't do is:
1. allow redistribution without getting an additional license from D.M.
What I should also say is the back end was built without ever expecting
it to be open source, hence the comments are generally notes to myself.
It's pretty dense, not very attractive, and full of bad style and
detritus. However, it does work.
I also need to point out that in various places there are contributions
by Pat Nelson (initial elf work), Mike Cote and John Micco (inline
assembler). The instruction scheduler is based on ideas of Steve Russell's.
