completion.exp: what is "a64l"?

completion.exp: what is "a64l"?

[Michael Snyder]

Hi, 

The gdb test "completion.exp" is failing for me because 
it is expecting to see a symbol whose name begins with
"a64l" -- what is that?  There's no explanation for why
it should be expected to be there, or why its absence
amounts to a failure for gdb.

Michael

Re: completion.exp: what is "a64l"?

[Daniel Berlin]

Michael Snyder <msnyder@cygnus.com> writes:

> Hi, 
> 
> The gdb test "completion.exp" is failing for me because 
> it is expecting to see a symbol whose name begins with
> "a64l" -- what is that?  There's no explanation for why
> it should be expected to be there, or why its absence
> amounts to a failure for gdb.
> 
> Michael

It's part of glibc.
Don't ask me *what* it is, i just know what nm tells me.

nm /usr/lib/libc.a|grep a64
nm: s_isnanf.o: no symbols
nm: s_copysignf.o: no symbols
a64l.o:
0000000000000000 T a64l
0000000000000000 r a64l_table
0000000000004a64 t group_number

-- 
"The Stones, I love the Stones.  I watch them whenever I can.
Fred, Barney...
"-Steven Wright

Re: completion.exp: what is "a64l"?

[Andrew Cagney]

``a64l, l64a, l64a_r - convert between a long integer and a base-64 
ASCII string''

	Andrew

Re: completion.exp: what is "a64l"?

[Elena Zannoni]

Michael Snyder writes:
 > Hi, 
 > 
 > The gdb test "completion.exp" is failing for me because 
 > it is expecting to see a symbol whose name begins with
 > "a64l" -- what is that?  There's no explanation for why
 > it should be expected to be there, or why its absence
 > amounts to a failure for gdb.
 > 
 > Michael

I was working on solaris at the time, so maybe that's what happens there.
Frankly I don't remember. :-(

I am rebuilding a solaris gdb at the moment.

.....

Ok, I see what I was trying to do:

The output of that test from gdb is a list of possible completions.
I was trying to match on the last line of the output (on solaris):

p 'a
a0		a7	  acosh		alarm	    ascftime	 atexit
a1		a8	  add		ald	    asctime	 atof
a10		a9	  add_concell  aldst	    asctime_r	 atoi
a11		aT	  adddays	allocid	    asin	 atol
a2		abort	  additem	allocstatic  asinh	 atoll
a3		abs	  addsev	almhdlr	    atan	 awake
a4		access	  addseverity  altzone	    atan2	
a5		acct	  adjtime	append	    atanh	 
a6		acl	  adjustwcptr  argc	    atanhi	 
a64l		acos	  advance	argv	    atanlo	 
(gdb) p 'a


So yes, my bad. Feel free to change it. Making sure we get some of the
'a' symbols in break.c before the gdb prompt is probably good enough,
those symbols should always be found.

Elena

Re: completion.exp: what is "a64l"?

[Andrew Cagney]

> a64l		acos	  advance	argv	    atanlo	 
> (gdb) p 'a
> 
> 
> So yes, my bad. Feel free to change it. Making sure we get some of the
> 'a' symbols in break.c before the gdb prompt is probably good enough,
> those symbols should always be found.


Should the C program be changed to rig the results:

(gdb) p 'aaa<tab>
aaabuilders aaaescorts aaamovers

	Andrew

Re: completion.exp: what is "a64l"?

[Michael Snyder]

Andrew Cagney wrote:
> 
> ``a64l, l64a, l64a_r - convert between a long integer and a base-64
> ASCII string''

Is there any reason why a GDB test for completion
should require the presence of this symbol?

Re: completion.exp: what is "a64l"?

[Elena Zannoni]

Michael Snyder writes:
 > Andrew Cagney wrote:
 > > 
 > > ``a64l, l64a, l64a_r - convert between a long integer and a base-64
 > > ASCII string''
 > 
 > Is there any reason why a GDB test for completion
 > should require the presence of this symbol?

No, read further in the thread.

You can even change the completion test. Pick another letter that is
not 'a' and that makes more sense for the file we are debugging.

Elena

Re: completion.exp: what is "a64l"?

[Michael Snyder]

Elena Zannoni wrote:
> 
> Michael Snyder writes:
>  > Andrew Cagney wrote:
>  > >
>  > > ``a64l, l64a, l64a_r - convert between a long integer and a base-64
>  > > ASCII string''
>  >
>  > Is there any reason why a GDB test for completion
>  > should require the presence of this symbol?
> 
> No, read further in the thread.
> 
> You can even change the completion test. Pick another letter that is
> not 'a' and that makes more sense for the file we are debugging.

Why don't I just take the literal "a64l" out of the expect target?