[Bug symtab/15103] New: disassemble file:function should work

[Bug symtab/15103] New: disassemble file:function should work

[dje at google dot com]

http://sourceware.org/bugzilla/show_bug.cgi?id=15103

             Bug #: 15103
           Summary: disassemble file:function should work
           Product: gdb
           Version: HEAD
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: symtab
        AssignedTo: unassigned@sourceware.org
        ReportedBy: dje@google.com
    Classification: Unclassified


bash$ gdb gdb
(gdb) b parse_number
Breakpoint 3 at 0x53793e: parse_number. (7 locations)
(top-gdb) i b
Num     Type           Disp Enb Address            What
1       breakpoint     keep y   0x00000000006edb0c in internal_error
                                                   at ../../src/gdb/utils.c:956
2       breakpoint     keep y   0x00000000004ed3e8 in info_command
                                                   at
../../src/gdb/cli/cli-cmds.c:221
        silent
        return
3       breakpoint     keep y   <MULTIPLE>
3.1                         y     0x000000000053793e in parse_number
                                                   at
../../src/gdb/cp-name-parser.y:1360
3.2                         y     0x000000000054c986 in parse_number
                                                   at ../../src/gdb/p-exp.y:861
3.3                         y     0x0000000000549dd8 in parse_number
                                                   at
../../src/gdb/m2-exp.y:665
3.4                         y     0x00000000005469cc in parse_number
                                                   at
../../src/gdb/go-exp.y:711
3.5                         y     0x00000000005442bc in parse_number
                                                   at ../../src/gdb/f-exp.y:674
3.6                         y     0x00000000005414a8 in parse_number
                                                   at
../../src/gdb/jv-exp.y:701
3.7                         y     0x0000000000531185 in parse_number
                                                   at
../../src/gdb/c-exp.y:1654
(top-gdb) disas p-exp.y:parse_number
No symbol "p" in current context.
(top-gdb) b p-exp.y:parse_number
Note: breakpoint 3 also set at pc 0x54c986.
Breakpoint 4 at 0x54c986: file ../../src/gdb/p-exp.y, line 861.
(top-gdb) 

If "b p-exp.y:parse_number" works,
why doesn't "disas p-exp.y:parse_number" ?

-- 
Configure bugmail: http://sourceware.org/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

[Bug symtab/15103] disassemble file:function should work

[tromey at redhat dot com]

http://sourceware.org/bugzilla/show_bug.cgi?id=15103

Tom Tromey <tromey at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |tromey at redhat dot com

--- Comment #1 from Tom Tromey <tromey at redhat dot com> 2013-02-06 12:18:03 UTC ---
(In reply to comment #0)

> If "b p-exp.y:parse_number" works,
> why doesn't "disas p-exp.y:parse_number" ?

disassemble takes expressions, not linespecs.

-- 
Configure bugmail: http://sourceware.org/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

[Bug symtab/15103] disassemble file:function should work

[dje at google dot com]

http://sourceware.org/bugzilla/show_bug.cgi?id=15103

--- Comment #2 from dje at google dot com 2013-02-06 18:19:18 UTC ---
Re: disassemble takes expressions, not linespecs.

That's not really my point though, is it.

Let's take a step back.
How do I explain to a user that while, yes, s/he can set a breakpoint on
p-exp.y:parse_number, s/he cannot disassemble it (in any easy way).
I could be missing something of course.

btw, somewhat related and somewhat tangential, I think we'll want to expand
expressions to take file names.
How do I (easily) print foo.c:baz instead of bar.c:baz from an arbitrary point
in the program?

"print foo.c:baz" is the obvious choice of course, but I if necessary (and
acceptable) I could see "print -f foo.c baz". It does preclude expressions
based on multiple files, but I can live with that.
[Some might want /X instead of -f, where X is some as yet unused display option
for print, but I'm not convinced it's better.]
[Again, this is all assuming "print foo.c:baz" can't work.
Plus, while linespecs don't always rquire filenames to be quoted, I could live
with "print 'foo.c':baz" too, if necessary.]

But getting back to disassemble,
from a u/i perspective I can't rationalize the current behaviour.

-- 
Configure bugmail: http://sourceware.org/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

[Bug symtab/15103] disassemble file:function should work

[dje at google dot com]

http://sourceware.org/bugzilla/show_bug.cgi?id=15103

--- Comment #3 from dje at google dot com 2013-02-06 18:23:02 UTC ---
btw, for completeness sake, and all that :-)

I realize one can do "info line p-exp.y:parse_number", examine the output,
and pass that to disassemble.  But it's a hack, not a solution.

-- 
Configure bugmail: http://sourceware.org/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

[Bug symtab/15103] disassemble file:function should work

[tromey at redhat dot com]

http://sourceware.org/bugzilla/show_bug.cgi?id=15103

--- Comment #4 from Tom Tromey <tromey at redhat dot com> 2013-02-06 18:29:23 UTC ---
(In reply to comment #2)
> Re: disassemble takes expressions, not linespecs.
> 
> That's not really my point though, is it.

I didn't know.  I thought it was an actual question.

> Let's take a step back.
> How do I explain to a user that while, yes, s/he can set a breakpoint on
> p-exp.y:parse_number, s/he cannot disassemble it (in any easy way).
> I could be missing something of course.
> 
> btw, somewhat related and somewhat tangential, I think we'll want to expand
> expressions to take file names.
> How do I (easily) print foo.c:baz instead of bar.c:baz from an arbitrary point
> in the program?

There's already an extension for this.
Try:  disassemble 'p-exp.y'::parse_number

> But getting back to disassemble,
> from a u/i perspective I can't rationalize the current behaviour.

Linespecs might have been a better choice, but the choice has already been
made.

-- 
Configure bugmail: http://sourceware.org/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

[Bug symtab/15103] disassemble file:function should work

[dje at google dot com]

http://sourceware.org/bugzilla/show_bug.cgi?id=15103

--- Comment #5 from dje at google dot com 2013-02-06 19:16:18 UTC ---
Blech.  Silly me for applying sane rules of design to gdb.

It's break foo.c:bar, but it's disassemble 'foo.c'::bar.

Except that disassemble 'foo.c'::bar doesn't work if the current language is
c++,
which is what my testcase was written in when I checked whether it worked
[filed as http://sourceware.org/bugzilla/show_bug.cgi?id=15109]

I consider this bug still open though. :-)
At least until a compelling reason is given for why neither of the following
can be done:
1) allow :: in linespecs
2) allow : in disassemble, et.al.

-- 
Configure bugmail: http://sourceware.org/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

[Bug symtab/15103] disassemble file:function should work

[tromey at sourceware dot org]

https://sourceware.org/bugzilla/show_bug.cgi?id=15103

Tom Tromey <tromey at sourceware dot org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |tromey at sourceware dot org

--- Comment #6 from Tom Tromey <tromey at sourceware dot org> ---
I found another spot where it would be better for disassemble
to take a linespec rather than an expression.  Like Pascal, in 
Ada a call of a parameter-less function does not require parens.
So you can end up with confusing output like:

    (gdb) disassemble Mult_Float
    Too few arguments in function call.

Oops, this is trying to call a function; you have to stick an "&"
or "'address" in there.

-- 
You are receiving this mail because:
You are on the CC list for the bug.