Completion on static functions

Completion on static functions

[Eli Zaretskii]

Sometimes "break foo TAB" does not complete the name of a function,
even though typing its full name as in "break foo_bar RET" sets a
breakpoint without any complaints.  It happens with static functions,
but with all of them.  Perhaps only inlined functions?

I see this both in 7.5.1 and in 7.5.91, on MS-Windows, with programs
compiled with MinGW GCC 4.7.2.

Does anyone else see this?  Can this annoyance be fixed somehow?

TIA

Re: Completion on static functions

[Eli Zaretskii]

> Date: Sun, 07 Apr 2013 19:28:16 +0300
> From: Eli Zaretskii <eliz@gnu.org>
> 
> Sometimes "break foo TAB" does not complete the name of a function,
> even though typing its full name as in "break foo_bar RET" sets a
> breakpoint without any complaints.  It happens with static functions,
> but with all of them.  Perhaps only inlined functions?
  ^^^^^^^^^^^^^^^^^^^^
"but NOT with all of them", of course.

Re: Completion on static functions

[Jan Kratochvil]

On Fri, 12 Apr 2013 08:19:22 +0200, Eli Zaretskii wrote:
> So does this silence mean that no one else bumped into this yet?

Maybe nobody has noticed, primarily there is no reproducer above.


Regards,
Jan

Re: Completion on static functions

[Markus Teich]

You could try creating a variable object out of the functionname.
As far as i can tell, that (only?) fails, if the function is inlined.

--Markus


Am 12.04.2013 08:47, schrieb Jan Kratochvil:
> On Fri, 12 Apr 2013 08:19:22 +0200, Eli Zaretskii wrote:
>> So does this silence mean that no one else bumped into this yet?
> 
> Maybe nobody has noticed, primarily there is no reproducer above.
> 
> 
> Regards,
> Jan
> 

Re: Completion on static functions

[Eli Zaretskii]

> Date: Fri, 12 Apr 2013 08:47:49 +0200
> From: Jan Kratochvil <jan.kratochvil@redhat.com>
> Cc: gdb@sourceware.org
> 
> On Fri, 12 Apr 2013 08:19:22 +0200, Eli Zaretskii wrote:
> > So does this silence mean that no one else bumped into this yet?
> 
> there is no reproducer above.

Sorry about that.  I didn't imagine the problem was so rare.  My bad.

Try this: build the latest sharutils 4.13.4 with CFLAGS='-g3 -O2',
then do this:

  $ cd src && gdb ./shar
  (gdb) break generate_full_ TAB

(There is a static function named generate_full_header in shar.c.)

As mentioned before, the problem happened to me with GCC 4.7.2 on
MinGW.  As expected, GDB reports (in "info source") that the debug
info is DWARF 2 with preprocessor macro info.

Let me know if you need more info from GDB, I have the build directory
handy with the object and executable files that trigger this problem.

Re: Completion on static functions

[Markus Teich]

Try compiling with -O0 to see if these functions indeed get inlined.

--Markus


Am 12.04.2013 10:12, schrieb Eli Zaretskii:
>> Date: Fri, 12 Apr 2013 08:47:49 +0200
>> From: Jan Kratochvil <jan.kratochvil@redhat.com>
>> Cc: gdb@sourceware.org
>>
>> On Fri, 12 Apr 2013 08:19:22 +0200, Eli Zaretskii wrote:
>>> So does this silence mean that no one else bumped into this yet?
>>
>> there is no reproducer above.
> 
> Sorry about that.  I didn't imagine the problem was so rare.  My bad.
> 
> Try this: build the latest sharutils 4.13.4 with CFLAGS='-g3 -O2',
> then do this:
> 
>   $ cd src && gdb ./shar
>   (gdb) break generate_full_ TAB
> 
> (There is a static function named generate_full_header in shar.c.)
> 
> As mentioned before, the problem happened to me with GCC 4.7.2 on
> MinGW.  As expected, GDB reports (in "info source") that the debug
> info is DWARF 2 with preprocessor macro info.
> 
> Let me know if you need more info from GDB, I have the build directory
> handy with the object and executable files that trigger this problem.
> 

Re: Completion on static functions

[Eli Zaretskii]

> Date: Fri, 12 Apr 2013 10:20:50 +0200
> From: Markus Teich <markus.teich@stusta.mhn.de>
> 
> Try compiling with -O0 to see if these functions indeed get inlined.

No need to recompile:

  (gdb) info address print_header_stamp
  Symbol "print_header_stamp" is a function at address 0x401cec.
  (gdb) info address main
  Symbol "main" is a function at address 0x427b10.

But:

  (gdb) info address generate_full_header
  No symbol "generate_full_header" in current context.

And if I disassemble the code, I don't see any call instructions for
that function (nor even its caller, configure_shar, and its caller's
caller, initialize).

So I think it's quite clear that this function is inlined, as are many
others in that program.