[v7.1,3/8] Use xml-syscall to compare syscall numbers in arm_linux_sigreturn_return-addr

  In this v7.1:

Use xml_get_syscall_number to return the number and compare with that.

Note the subsequent patches in the series have been updated with that change.

---

This patch changes checks for sigreturn and rt_sigreturn syscalls in
arm-linux-tdep.c from magic numbers to numbers computed from
syscall/arm-linux.xml.

It also adds a new function to xml-syscall.h/c to get the syscall number
called get_syscall_number.

No regressions, tested on ubuntu 14.04 ARMv7 and x86.
With gdbserver-{native,extended} / { -marm -mthumb }

gdb/ChangeLog:

	* arm-linux-tdep.c (arm_linux_sigreturn_return_addr): Use
	get_syscall_number.
	* xml-syscall.c (get_syscall_number): New function.
	* xml-syscall.h (get_syscall_number): New declaration.
---
 gdb/arm-linux-tdep.c | 5 ++++-
 gdb/xml-syscall.c    | 8 ++++++++
 gdb/xml-syscall.h    | 6 ++++++
 3 files changed, 18 insertions(+), 1 deletion(-)
  

Antoine Tremblay <antoine.tremblay@ericsson.com> writes:

> +int get_syscall_number (struct gdbarch *gdbarch,
> +			const char *syscall_name)

get_syscall_number should be written in the next line.

> +{
> +  init_syscalls_info (gdbarch);
> +
> +  return xml_get_syscall_number (gdbarch, syscall_name);
> +}

Patch is good to me otherwise.
  

You also need a dummy version of the function under

 #ifndef HAVE_LIBEXPAT

in order to avoid breaking the build when expat is not available.

BTW, I think this patch now makes libexpat a hard requirement
for basic ARM Linux debugging.

Thanks,
Pedro Alves
  

On 12/17/2015 05:17 AM, Pedro Alves wrote:
>
> You also need a dummy version of the function under
>
>   #ifndef HAVE_LIBEXPAT
>
> in order to avoid breaking the build when expat is not available.
>

Right indeed thx.

> BTW, I think this patch now makes libexpat a hard requirement
> for basic ARM Linux debugging.
>

OK I had not considered that at first, that sounds wrong to me compared 
to the option of having a #define and no hard requirement on expat.

On the other hand expat is needed for debugging with GDBServer...

Yao, had you considered this ?

Is it not better to have a #define considering this ?

Thanks,
Antoine
  

Antoine Tremblay <antoine.tremblay@ericsson.com> writes:

> On the other hand expat is needed for debugging with GDBServer...
>

Not really, we can build gdb with expat disabled, and it can still talks
with GDBserver.

> Yao, had you considered this ?
>
> Is it not better to have a #define considering this ?

Looks it is not easy to use xml-syscall for syscall numbers, probably we
can roll back to define these syscall number in macro and avoid using
xml things, like

#define ARM_SIGRETURN 119
  

On 12/17/2015 08:24 AM, Yao Qi wrote:
> Antoine Tremblay <antoine.tremblay@ericsson.com> writes:
>
>> On the other hand expat is needed for debugging with GDBServer...
>>
>
> Not really, we can build gdb with expat disabled, and it can still talks
> with GDBserver.
>

Humm really ? I thought it failed to read the target description and 
failed with errors like :

warning: while parsing target description (at line 10): Target 
description specified unknown architecture "arm"
warning: Could not load XML target description; ignoring
Remote 'g' packet reply is too long: 0...


>> Yao, had you considered this ?
>>
>> Is it not better to have a #define considering this ?
>
> Looks it is not easy to use xml-syscall for syscall numbers, probably we
> can roll back to define these syscall number in macro and avoid using
> xml things, like
>
> #define ARM_SIGRETURN 119
>

Yes I'll repost with that, I'll keep the __NR_sigreturn macros for 
GDBServer however.

Thanks,
Antoine