Re: [commit] Improved linker-debugger interface

[Sergio Durigan Junior <sergiodj@redhat.com>]

On Tuesday, June 25 2013, Joel Brobecker wrote:

>> I just committed the improved linker-debugger interface patch series
>> as attached.  For those of you using the patches in distributions,
>> some of the attached files have different filenames than previously
>> posted:
>
> I am seeing some issues on ia64-linux (rhES5):
>
>   | (gdb) start
>   | Temporary breakpoint 1 at 0x40000000000005a1: file bar.c, line 4.
>   | Starting program: /saco.a/users/brobecke/ex/simple/bar 
>   | warning: Probes-based dynamic linker interface failed.
>   | Reverting to original interface.
>   |
>   | Unknown numeric token on expression `0 8@r35'.
>
> I will investigate the problem with the unknown numeric token, but
> any recommendation before I start?

Hi Joel,

Thanks for the report.  There are two kinds of things happening here.
First, the current SDT parser (implemented on stap-probe.c) is not
recognizing one of the probe's arguments.  This is where the "Unknown
numeric token" comes from.  The initial "0" is making the parser
confused, and to be honest it is also making me confused :-).  I did not
implement the support for IA-64, FWIW.

I am looking into this issue right away.

> Also, I am wondering we really want a warning in this case -
> I think this is going to make the average user think that there
> is something wrong and therefore that needs to be fixed. What
> do people think?

The second thing is the warning.  It is issues by Gary's patch, and I
agree that it could confuse users.  However, I think it is still a good
thing to have.  Maybe one could create a debug flag that would enable
printing such warnings?  Just an idea.

I will get back to you when I have more news about what's going on with
the SDT parser.

Thanks,

-- 
Sergio