On Wednesday 26 June 2002 h:43, Kevin D. Clark wrote: > Tom Rauschenbach <[EMAIL PROTECTED]> writes: > > I'd really like to see the ability to have *tiny* pages to give extra > > hardware support to debuggers. Imagine getting an interupt when a > > certain variable is changed > > You know, of course, that most debuggers allow you to catch reads and > writes to a certain memory location already, right?
Sure, but they usually do it by inserting an illegal instruction at the beginning of each statement boundary (there's a proper phrase for that in C but it escapes me now) and the overhead of catching that interupt and checking if the location has changed is huge. Typically the program being debugged runs several orders of magnitude slower. In some applications this even changes the behavior. I've used electric fence. Because it basically puts every malloced block in its own page with an unmapped page above and below it, it gives the effect I'm talking about but with huge memory overhead. One can't always use it. Sometimes one runs out of available space before the problem occurs. It's great when it works though, and there's very little performance impact. But hardware support for debuggers would be hard to beat. > > Sometimes this is very handy... > > > , or when a malloced block overflows. > > ElectricFence (and similar tools) are useful here. > > --kevin ***************************************************************** To unsubscribe from this list, send mail to [EMAIL PROTECTED] with the text 'unsubscribe gnhlug' in the message body. *****************************************************************
