This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [PATCH] Fix length calculation in aarch64_linux_set_debug_regs
- From: Pedro Alves <palves at redhat dot com>
- To: Simon Marchi <simon dot marchi at ericsson dot com>, gdb-patches at sourceware dot org
- Cc: qiyaoltc at gmail dot com
- Date: Mon, 02 Nov 2015 16:00:04 +0000
- Subject: Re: [PATCH] Fix length calculation in aarch64_linux_set_debug_regs
- Authentication-results: sourceware.org; auth=none
- References: <1446475684-31936-1-git-send-email-simon dot marchi at ericsson dot com>
On 11/02/2015 02:48 PM, Simon Marchi wrote:
> There is this build failure when building in C++:
>
> In file included from build-gnulib/import/stddef.h:45:0,
> from /usr/include/time.h:37,
> from build-gnulib/import/time.h:41,
> from build-gnulib/import/sys/stat.h:44,
> from ../bfd/bfd.h:44,
> from /home/simark/src/binutils-gdb/gdb/common/common-types.h:35,
> from /home/simark/src/binutils-gdb/gdb/common/common-defs.h:44,
> from /home/simark/src/binutils-gdb/gdb/nat/aarch64-linux-hw-point.c:19:
> /home/simark/src/binutils-gdb/gdb/nat/aarch64-linux-hw-point.c: In function âvoid aarch64_linux_set_debug_regs(const aarch64_debug_reg_state*, int, int)â:
> /home/simark/src/binutils-gdb/gdb/nat/aarch64-linux-hw-point.c:564:64: error: âcountâ cannot appear in a constant-expression
> iov.iov_len = (offsetof (struct user_hwdebug_state, dbg_regs[count - 1])
> ^
>
> I don't really understand the length computation done here.
>
> From what I understand, the dbg_regs array in the user_hwdebug_state structure
> is 16 elements long, but we don't use all of them. We want iov_len to reflect
> only the used bytes. If that's true, I don't think that the current code is
> correct.
> Instead, it can be computed simply with:
>
> offsetof (struct user_hwdebug_state, dbg_regs) + count * sizeof (regs.dbg_reg[0]);
>
> Does it make sense?
IIUYC, you're pointing out two issues:
#1 - the offsetof that doesn't work in C++.
#2 - an off-by-one.
I don't know enough about Aarch64 to judge #1, but it does sound right to me.
On #2, I saw the same on x86. See my fix here:
https://sourceware.org/ml/gdb-patches/2015-02/msg00213.html
I think it's a little nicer to hide away the offsetof+sizeof.
Thanks,
Pedro Alves