Re: PR/15448: add a compiler timestamp to PCH files (take 2)

[Geoff Keating <geoffk at geoffk dot org>]

On 17/02/2005, at 5:31 AM, Paul Brook wrote:

> On Thursday 17 February 2005 13:26, Paolo Bonzini wrote:
> > > I think this solves the problem nicely, and fixes the bootstrap
> > > problem.  The only other issue I can think of, is, it is critical 
> > > (for
> > > the installation of the libstdc++ gch file) that the install process
> > > leaves the mtime alone on the compiler binary.  touch -r might be
> > > necessary to make it work.  touch -r isn't very portable.
> >
> > The PCH would need to be rebuilt upon installation, or the st_size 
> > could
> > be used alone.  The approach of checking addresses are unreliable
> > because of PIE.
>
> Is the size reliable? It's quite common for binaries to be stripped 
> during
> installation. The only thing I can thing of that should be unmodified 
> is the
> *relative* addresses of certain key functions.

PCH currently assumes that the *absolute* addresses stay constant.

Attachment: smime.p7s
Description: S/MIME cryptographic signature