This is the mail archive of the
mailing list for the binutils project.
Re: dwarf2: pointer size vs address size
- From: Nick Clifton <nickc at redhat dot com>
- To: DJ Delorie <dj at redhat dot com>
- Cc: binutils at sourceware dot org
- Date: Wed, 28 Jun 2006 10:19:12 +0100
- Subject: Re: dwarf2: pointer size vs address size
- References: <200606261846.k5QIkQx8029367@greed.delorie.com>
In various places, we refer to "pointer size" to refer to things the
dwarf2 spec calls "address size". Example: include/elf/dwarf.c has
cu_pointer_size, and readelf prints "Pointer size:" for the address
size given in compilation unit headers.
On some platforms, the pointer size is not the same as the address
size. Example: m16c has a 20 bit address space, but only 16 bit
pointers, and we use 32 bit addresses in its dwarf information.
To avoid confusion, would anyone object to renaming those things to
say "address" instead of "pointer" ?
No objections, in fact definite encouragement.
On a related note, readelf ignores this field when printing EH tables.
It blindly assumes that EH addresses will be either 32 or 64 bit, and
uses the machine type to decode it. Why?
Poor coding assumption.