This is the mail archive of the
mailing list for the GCC project.
Re: DECL_SIZE issues (was: PR 6212)
- From: kenner at vlsi1 dot ultra dot nyu dot edu (Richard Kenner)
- To: rth at redhat dot com
- Cc: gcc at gcc dot gnu dot org
- Date: Fri, 17 May 02 08:40:14 EDT
- Subject: Re: DECL_SIZE issues (was: PR 6212)
> I saw the claim, but not the explanation of why.
At minimum, PRs 2511, 3325, 3326, 3347, 5593.
At the heart of all, we mis-promote bit-fields.
This is a different issue and is talking about the case where DECL_SIZE
is *smaller* than TYPE_PRECISION for integral types. I agree that's wrong,
but that's not the cases I'm talking about.
First of all, the most common cases I'm referering to are where the type is
a record or array type. Secondly, I'm talking about cases where DECL_SIZE
would be *larger* than the useful data in the type (not necessarily larger
than TYPE_SIZE if the type was a record tye that was padded by alignment
or an integral type with TYPE_PRECISION less than TYPE_SIZE).
I still haven't heard an explanation of why we shouldn't pad the object in
question simply by making DECL_SIZE different from TYPE_SIZE.