Re: [avr-gcc-list] atmega32

[Erik Christiansen]

On Tue, Jun 10, 2003 at 11:48:50AM +0200, David Brown wrote:
> > address@hidden wrote:
> >
> > >> After compiling I get a "clock skew detected, your build
> > > may be incomplete"
> > >> warning, but it seems to run fine on the chips.
> >
> > > IIRC, that comes from using Win 95/98 where their timer resolution
> > > is coarse.
> >
> > IMHO, it's the time stamp resolution of the FAT filesystem which has a
> > granularity of only 2 seconds.
> >
> 
> Ah, that sounds like the correct explanation.  I use NTFS on NT/W2K
> machines, and FAT32 on Win9x machines.  I assumed it was the OS's file
> handling that was limited rather than the underlying file system.

   Clock skew can be detected, even on unix networks, if sharing build
   files between hosts, and the two (time & date) clocks are not
   synchronised. When the compiling host lags the fileserver, the
   generated files are "older" than their inputs, which is a logical
   impossibility. This makes make's decision making erroneous, hence the
   warning. (In practice, it may just check that input file timestamps
   are in the future, since this is a quicker test for the same effect.)

   When this last happened here, a few years back, I synchronised date &
   time, effecting an immediate cure.

   It is not easy to imagine a way for this to happen on a single host,
   without something being seriously broken. Hopefully this is not your
   scenario. :-)

Regards,
Erik