[gclist] Re: gclist-digest V1 #87

[Charles Fiterman]

>    If compilers never changed or needed to be extended and
>    were known to be free of errors and completely reliable,
>    then we could claim that compilers were a general method
>    for translating high-level languages into correct machine
>    code, but in fact compilers do need to be changed and
>    extended and are prone to bugs.
>
>Like compilers, garbage collectors are hard to write, to debug,
>to port, and to extend.  Once upon a time I was responsible for
>a garbage collector that was used by thousands of programmers.
>It made a lot more sense for me to spend the time it took to get
>it right on a new processor or system than for those thousands
>of programmers to port their manual memory management code for
>thousands of programs.

We maintain our Garbage Collector on a number of compilers and
say "There is no such thing as a benign compiler change." Somebody
changes their compiler and at the very least we have to recompile
and send out new libraries. At the bad end we have weeks of work.

But I think this is special to commercial conservative collectors
which must replace every memory service of the compiler in a
reasonable way. If they have an _heapwalk we must have an _heapwalk
and they must be compatible. Doing this is the difference between
a toy system suitable for educational use and a commercial system.

Charles Fiterman	Geodesic Systems
Phone 312-728-7196	4745 N. Ravenswood
Fax   312-726-6096	Chicago Il 60015