This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [PATCH 0/3] Skip thumb call-indirect trampoline
- From: Will Newton <will dot newton at linaro dot org>
- To: Yao Qi <yao at codesourcery dot com>
- Cc: "gdb-patches at sourceware dot org" <gdb-patches at sourceware dot org>
- Date: Mon, 9 Jun 2014 08:48:34 +0100
- Subject: Re: [PATCH 0/3] Skip thumb call-indirect trampoline
- Authentication-results: sourceware.org; auth=none
- References: <1402279983-1907-1-git-send-email-yao at codesourcery dot com>
On 9 June 2014 03:13, Yao Qi <yao@codesourcery.com> wrote:
> We find the following test failures on arm targets with compilation
> flags '-mthumb -march=armv4t':
>
> FAIL: gdb.base/funcargs.exp: backtrace through call with trampolines
> FAIL: gdb.base/funcargs.exp: stepping back to main from function called with trampolines
> FAIL: gdb.base/funcargs.exp: stepping into function called with trampolines
> FAIL: gdb.base/funcargs.exp: stepping into indirectly called function
>
> These fails are caused by a gdb unrecognized trampoline added by gcc
> in this patch (written by Richard Earnshaw),
>
> [PATCH] ARM: Handle Thumb call-indirect in an EABI-compliant manner
> https://gcc.gnu.org/ml/gcc-patches/2005-01/msg00813.html
>
> This patch series is about teaching GDB to detect this new trampoline
> sequence. Note that Richard Earnshaw pointed out in the mail that
> his patch causes two GDB regressions, but I am unsure they are the
> ones I am fixing today.
>
> Patches are regression tested on both arm-none-eabi and
> arm-none-linux-gnu-eabi with various multi-libs. Fails above are
> fixed and no regressions.
>
> IMO, it is better for 7.8 release to pick up this fix.
>
> *** BLURB HERE ***
>
> Yao Qi (3):
> Skip 'bx reg' trampoline on arm-none-eabi
> Apply stub unwinder to 'bx reg' trampoline
> Skip 'bx reg' on arm-linux
>
> gdb/arm-linux-tdep.c | 15 +++++++++-
> gdb/arm-tdep.c | 79 ++++++++++++++++++++++++++++++++++++++++++++++++++--
> 2 files changed, 91 insertions(+), 3 deletions(-)
This series looks good to me, although I am not a maintainer.
--
Will Newton
Toolchain Working Group, Linaro