Re: [PATCH] bfd: Use size_t for length argument totarget_read_memory function passed into bfd_elf_bfd_from_remote_memory

[Jan Kratochvil <jan.kratochvil@redhat.com>]

On Fri, 01 Jun 2012 20:24:14 +0200, Hans-Peter Nilsson wrote:
> There's always bfd_size_type, though I haven't checked if it
> fits your needs.

bfd_size_type seems to be always >= size_t so it should work.


> For reference, the m32r-elf fail:

I can confirm it with
	./configure --target=m32r-elf; make


On Fri, 01 Jun 2012 21:54:10 +0200, Siddhesh Poyarekar wrote:
> On Fri, 1 Jun 2012 20:24:14 +0200, Hans-Peter wrote:
> > There's always bfd_size_type, though I haven't checked if it
> > fits your needs.
> 
> Should be OK as long as it always matches the size_t definition in the
> stddef.h that gcc ships.

It is larger in some cases than size_t; but GDB can adapt, it is still better
than it was smaller before.


> They're both typically unsigned long, but if
> that is so, then bfd_size_type should have been typedef'd to size_t
> anyway.

Unfortunately bfd_size_type is not always type-compatible with size_t so
passing pointers to prototyped functions would not work.


> > Should bfd.h include sysdep.h or what is missing?

sysdep.h inclusion has caused problems which were avoided by
	Re: recent change broke gdb build
	http://sourceware.org/ml/binutils/2012-05/msg00224.html
but they can be hit in other cases like here, so I do not think bfd.h can
include sysdep.h.


> An stddef.h include in bfd.h should fix this.

stddef.h may not exist on the host system, it should include sysdep.h instead.


> That or I can fix my patch to use bfd_size_type provided its size is always
> equal to size_t, so that the include is not needed. Which way would be
> preferable?

I think the bfd/ part should use bfd_size_type and the gdb/ part should
continue to use size_t where possible, only in some few cases it needs to be
prototype-compatible with bfd/ it would use bfd_size_type.

Going to post a patch today as keeping HEAD broken is not good.


Thanks,
Jan