compiler rebuild invalidates precompiled header

[Mike Stump]

On Monday, May 17, 2004, at 02:05 PM, Geoff Keating wrote:
I thought of that at the time.  The problem is that it's too
sensitive; modification times can change for many reasons (like NFS
time skew, or copying a file from one place to another).
NFS time skew cannot change a modification time of a file, honest.  
However, your point remains, doing this leads to false positives, and 
that is bad.

Maybe there is another change that is easier, maybe if we grabbed the 
address of a piece of data near the end of the data segment (akin to 
_edata) and validated it, that would be enough to catch a large enough 
class of problems for this issue to go away.  Advantage, no false 
positives.