This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Apple's implementation of precompiled headers
- To: Devang Patel <dpatel at apple dot com>
- Subject: Re: Apple's implementation of precompiled headers
- From: Daniel Egger <degger at fhm dot edu>
- Date: 04 Oct 2001 00:05:05 +0200
- Cc: Stan Shebs <shebs at apple dot com>, gcc at gcc dot gnu dot org
- References: <2BF9AC32-B76A-11D5-9AAF-00306558FC20@apple.com>
Am Die, 2001-10-02 um 21.17 schrieb 1002050270:
> Yes and No, because how/when to decide if it is 'changed' ?
> - content is changed
Yes.
> - content is same but different compiler options
> - content is same but different macro settings
Depends on the implementation. When complete "compiled" trees are
serialized this has to be taken into account. Otherwise, when the
macros are evaluated upon compilation, this is irrelevant.
> - content is same but different time stamp
> - content is same but different file system location
If you can ensure that the content is the same you can forget about
timestamps and location. But only if....
> Any suggestions ?
I can't think of a good implementation right now to check that the
content hasn't changed or is equal. Well, if the speedadvantage is
expected to come from the avoided parse and compilation process itself
then computing a hash from the headerfiles and storing it in the tree
might be a good idea.
--
Servus,
Daniel