This is the mail archive of the
gdb@sourceware.org
mailing list for the GDB project.
Re: 'maybe_software_singlestep' inserts breakpoints when record replay
- From: Hui Zhu <teawater at gmail dot com>
- To: Xingxing Pan <forandom at gmail dot com>
- Cc: gdb <gdb at sourceware dot org>
- Date: Wed, 10 Nov 2010 06:43:41 +0800
- Subject: Re: 'maybe_software_singlestep' inserts breakpoints when record replay
- References: <AANLkTi=fkBydf9Au2u3kHg4U2x0jt3H8jUqyKwBYhvK_@mail.gmail.com>
Are you work in trunk?
I remember I check in a patch to fix this issue before 7.2 release.
Thanks,
Hui
On Tue, Nov 9, 2010 at 23:40, Xingxing Pan <forandom@gmail.com> wrote:
> Hi guys,
>
> I'm porting the process record to an arch which only supports software
> singlestep.
>
> And I found that in the function 'resume', the function
> 'maybe_software_singlestep' inserts a software singlestep breakpoint
> although the inferior is in record replay, and return a zero value to
> the variable 'step'. When I use the command 'reverse-stepi', ?gdb
> cannot stop after reversely executing one instruction .
>
> I try to prevent 'maybe_software_singlestep' from inserting a software
> singlestep breakpoint when 'RECORD_IS_REPLAY' is asserted. Then the
> error is disappeared.
>
> I cannot figure out why we need to insert a software singlestep
> breakpoint when gdb is travelling along the record_list back and
> forth. Is this a bug?
>
> Thanks,
> Pan Xingxing
>