in_solib_call_trampoline

in_solib_call_trampoline

[Kris Warkentin]

Sorry for the dumb question but it's been a while since I brought my 
head branch sandbox up to date.  I'm sure I'm missing something obvious.

I was using set_gdbarch_in_solib_call_trampoline (gdbarch, 
in_plt_section) so that I could step into solibs, much like many other 
targets, but it looks like this method isn't supported anymore.  Can 
anyone tell me the accepted way to make gdb keep stepping past the plt 
section and into the solib?

cheers,

Kris

-- 
Stay up-to-date on all the QNX news!  Register at
http://www.qnx.com/news/forms/newsletter.html to
receive our newsletter.

Re: in_solib_call_trampoline

[Mark Kettenis]

   Date: Wed, 27 Jul 2005 15:26:57 -0400
   From: Kris Warkentin <kewarken@qnx.com>

   I was using set_gdbarch_in_solib_call_trampoline (gdbarch, 
   in_plt_section) so that I could step into solibs, much like many other 
   targets, but it looks like this method isn't supported anymore.  Can 
   anyone tell me the accepted way to make gdb keep stepping past the plt 
   section and into the solib?

I'm a bit confused here myself, but I have you taken a look at
skip_trampoline_code?

Mark

Re: in_solib_call_trampoline

[Kris Warkentin]

Well, I've been looking at the frame_unwind_sniffers...I guess I'll try 
that next.  Kind of annoying not being able to step into shared libs 
anymore. ;-)

cheers,

Kris

Mark Kettenis wrote:

>   Date: Wed, 27 Jul 2005 15:26:57 -0400
>   From: Kris Warkentin <kewarken@qnx.com>
>
>   I was using set_gdbarch_in_solib_call_trampoline (gdbarch, 
>   in_plt_section) so that I could step into solibs, much like many
>other 
>   targets, but it looks like this method isn't supported anymore.  Can 
>   anyone tell me the accepted way to make gdb keep stepping past the
>plt 
>   section and into the solib?
>
>I'm a bit confused here myself, but I have you taken a look at
>skip_trampoline_code?
>
>Mark
>  
>

-- 
Stay up-to-date on all the QNX news!  Register at
http://www.qnx.com/news/forms/newsletter.html to
receive our newsletter.

Re: in_solib_call_trampoline

[Mark Kettenis]

   Date: Wed, 27 Jul 2005 16:05:21 -0400
   From: Kris Warkentin <kewarken@qnx.com>

   Well, I've been looking at the frame_unwind_sniffers...I guess I'll try 
   that next.  Kind of annoying not being able to step into shared libs 
   anymore. ;-)

Having you target unwind from within .plt is a good idea anyway; your
program might end up there once in a while if it encounters a random
signal.

Mark

Re: in_solib_call_trampoline

[Daniel Jacobowitz]

On Wed, Jul 27, 2005 at 09:59:30PM +0200, Mark Kettenis wrote:
>    Date: Wed, 27 Jul 2005 15:26:57 -0400
>    From: Kris Warkentin <kewarken@qnx.com>
> 
>    I was using set_gdbarch_in_solib_call_trampoline (gdbarch, 
>    in_plt_section) so that I could step into solibs, much like many other 
>    targets, but it looks like this method isn't supported anymore.  Can 
>    anyone tell me the accepted way to make gdb keep stepping past the plt 
>    section and into the solib?
> 
> I'm a bit confused here myself, but I have you taken a look at
> skip_trampoline_code?

I thought that gdbarch_skip_solib_resolver and
in_solib_dynsym_resolve_code would be enough.  I'm not sure if all the
current skip_trampoline_code bits are necessary.

-- 
Daniel Jacobowitz
CodeSourcery, LLC

Re: in_solib_call_trampoline

[Randolph Chung]

>> I'm a bit confused here myself, but I have you taken a look at
>> skip_trampoline_code?
> 
> I thought that gdbarch_skip_solib_resolver and
> in_solib_dynsym_resolve_code would be enough.  I'm not sure if all the
> current skip_trampoline_code bits are necessary.

Well, on hppa skip_trampoline_code knows about quite a few other types
of trampolines in addition to solib PLT trampolines.

randolph
-- 
Randolph Chung
Debian GNU/Linux Developer, hppa/ia64 ports
http://www.tausq.org/

Re: in_solib_call_trampoline

[Kris Warkentin]

Well, on the bright side, it turns out there's no problem with solibs 
and backtraces.  It turns out that I'm an idiot and chased this problem 
all week only just now to notice that symbols weren't being loaded for 
the solibs.  So the problem is with auto-loading of symbols, not 
backtracing and solib support.

On the bright side, in looking at the linux support, I was able to rip a 
few things out of our tdep files and generally simplify things so it 
wasn't a complete waste of a week.  Still feel a bit of a dork though....;-)

Have a good weekend and thanks for the help.

cheers,

Kris

Randolph Chung wrote:

>>>I'm a bit confused here myself, but I have you taken a look at
>>>skip_trampoline_code?
>>>      
>>>
>>I thought that gdbarch_skip_solib_resolver and
>>in_solib_dynsym_resolve_code would be enough.  I'm not sure if all the
>>current skip_trampoline_code bits are necessary.
>>    
>>
>
>Well, on hppa skip_trampoline_code knows about quite a few other types
>of trampolines in addition to solib PLT trampolines.
>
>randolph
>  
>

-- 
Stay up-to-date on all the QNX news!  Register at
http://www.qnx.com/news/forms/newsletter.html to
receive our newsletter.