At 06:58 PM 9/23/00 +0000, you wrote:
>your patch solves the thread-unsafeness when two threads handle, at the same time,
>two different maps
>but, it doesn't work (still unsafe) when same map is manipulated
>basically, any manipulation of a FILE_VIEW shall be done with the cs locked...
>
>I agree that such a program would be poorly written, but the API shall be
>guarded against this
Probably. I made the assumption that no program could ever work reliably
by relying on the system not protecting its internal data structures; but to
do what you want it would be imperative to not do *more* than Windows
else there is the risk of breaking compatibility. It would need more tests.
There is the small matter of the available time, you see.
Gerard
