This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [PATCH][SH] Signal handler unwinding for SH-Linux
- From: Yao Qi <yao at codesourcery dot com>
- To: Chung-Lin Tang <cltang at codesourcery dot com>
- Cc: <gdb-patches at sourceware dot org>, Thomas Schwinge <thomas_schwinge at mentor dot com>
- Date: Sun, 29 Apr 2012 21:01:24 +0800
- Subject: Re: [PATCH][SH] Signal handler unwinding for SH-Linux
- References: <4F9CF631.9080807@codesourcery.com>
On 04/29/2012 04:05 PM, Chung-Lin Tang wrote:
> The test results show improvement on many sighandler related tests, with
> no regressions. Some gdb.base/sigbpt.exp tests change from FAIL->KFAIL,
> though I'm not completely sure what that indicates.
Chung-Lin,
Not sure the comment at the beginning of gdb.base/sigbpt.exp is relevant
to your KFAILs.
# This test is known to tickle the following problems: kernel letting
# the inferior execute both the system call, and the instruction
# following, when single-stepping a system call; kernel failing to
# propogate the single-step state when single-stepping the sigreturn
# system call, instead resuming the inferior at full speed; GDB
# doesn't know how to software single-step across a sigreturn
# instruction. Since the kernel problems can be "fixed" using
# software single-step this is KFAILed rather than XFAILed.
If this comment is related to your KFAILs, this can be handled by
computing the "next pc" of sigreturn or rt_sigreturn. Please have a
look at functions {arm,mips,tic6x}_linux_syscall_next_pc. You can do
something similar in SH backend. Hope it helps.
--
Yao (éå)