[Bug c++/14834] New: Crash while trying to read field

[Bug c++/14834] New: Crash while trying to read field

[icegood1980 at gmail dot com]

http://sourceware.org/bugzilla/show_bug.cgi?id=14834

             Bug #: 14834
           Summary: Crash while trying to read field
           Product: gdb
           Version: 7.5
            Status: NEW
          Severity: normal
          Priority: P2
         Component: c++
        AssignedTo: unassigned@sourceware.org
        ReportedBy: icegood1980@gmail.com
    Classification: Unclassified


Just typed 
(gdb) p this[0]
for instantiation of template class
Obtained
...<list of fields from base classes>, ..., maxDisp = 
/build/buildd/gdb-7.5/gdb/findvar.c:176: internal-error: extract_typed_address:
type is not a pointer or reference
A problem internal to GDB has been detected,
further debugging may prove unreliable.


where maxDisp (private, double) just seems to be 1st own field of that template
class.
Simply p maxDisp from command line is OK.
==========
version GNU gdb (GDB) 7.5-ubuntu

-- 
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 c++/14834] Crash while trying to read field

[tromey at redhat dot com]

http://sourceware.org/bugzilla/show_bug.cgi?id=14834

Tom Tromey <tromey at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
                 CC|                            |tromey at redhat dot com
         Resolution|                            |FIXED
   Target Milestone|---                         |7.6

--- Comment #1 from Tom Tromey <tromey at redhat dot com> 2012-11-12 21:38:17 UTC ---
Looks like https://bugzilla.redhat.com/show_bug.cgi?id=849357
This is fixed in CVS, so I'm closing.
If you try CVS head and it still fails for you, please reopen
with details.  Thanks.

-- 
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 c++/14834] Crash while trying to read field

[icegood1980 at gmail dot com]

http://sourceware.org/bugzilla/show_bug.cgi?id=14834

icegood <icegood1980 at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |REOPENED
         Resolution|FIXED                       |

--- Comment #2 from icegood <icegood1980 at gmail dot com> 2012-11-19 16:26:49 UTC ---
Unfortunately it's not seem to be a case. 

Problem  reproducable in 
GNU gdb (GDB) 7.5.0.20121112-cvs
in other context of program.


And again with double field. 
Cannot create simplified test model. Everything works under them. 

Log is:
{<McMove> = {<IoList> = {<Io> = {<BaseClass> = {<No data fields>}, _vptr.Io =
0x442ab0 <vtable for VolumeChangeMove<EnsembleNPT>+16>}, 
      static MaxNIoDefault = 100, ioList = 0x6809b0, maxNIo = 100, nIo = 1},
Name = "VolumeChangeMove", '\000' <repeats 63 times>, move_id = 0, nAttempt =
0, 
    nAccept = 0, sys = 0x7fffffffdb30, box = 0x659760, pairPotential =
0x65a890, bondPotential = 0x65cbb0, random = 0x659af0, FEnsemble = 0x659e70, 
    nPartType = @0x7fffffffdb78, nParticle = @0x7fffffffdb7c, cumul_time_proc =
0, cumul_time_sys = 0}, MaxRelativeLogChange = 
findvar.c:176: internal-error: extract_typed_address: type is not a pointer or
reference
A problem internal to GDB has been detected,
further debugging may prove unreliable.
Quit this debugging session? (y or n) 
MaxRelativeLogChange is double. I don't know yet what is wrong with field. It
read via scannf and nothing more "unusual". Valgring doesn't say smth wrong
with program. I don't know...

-- 
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 c++/14834] Crash while trying to read field

[tromey at redhat dot com]

http://sourceware.org/bugzilla/show_bug.cgi?id=14834

--- Comment #3 from Tom Tromey <tromey at redhat dot com> 2012-11-19 16:53:04 UTC ---
Can you get a stack trace from gdb at the point of failure?

-- 
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 c++/14834] Crash while trying to read field

[icegood1980 at gmail dot com]

http://sourceware.org/bugzilla/show_bug.cgi?id=14834

--- Comment #4 from icegood <icegood1980 at gmail dot com> 2012-11-19 17:19:17 UTC ---
tromey at redhat dot com <sourceware-bugzilla <at> sourceware.org> writes:

> 
> http://sourceware.org/bugzilla/show_bug.cgi?id=14834
> 
> --- Comment #3 from Tom Tromey <tromey at redhat dot com> 2012-11-19 16:53:04 UTC ---
> Can you get a stack trace from gdb at the point of failure?
> 

With my program everything is OK. "where" works and i can even continue
debugging.

But if you mean stack of gdb itself... I don't know how to do that. 

I applied debigging of  gdb under own instance but inner instance covers
terminal so "where" command applicable to inner instance, not outer.

-- 
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 c++/14834] Crash while trying to read field

[tromey at redhat dot com]

http://sourceware.org/bugzilla/show_bug.cgi?id=14834

--- Comment #5 from Tom Tromey <tromey at redhat dot com> 2012-11-19 17:21:58 UTC ---
(In reply to comment #4)

> 
> But if you mean stack of gdb itself... I don't know how to do that. 
>

When you get the "Quit this debugging session" prompt, in another
window, use "ps" to find the PID of gdb, then "gdb -p PID".
Then in the new gdb, "bt"

Thanks.

-- 
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 c++/14834] Crash while trying to read field

[icegood1980 at gmail dot com]

http://sourceware.org/bugzilla/show_bug.cgi?id=14834

--- Comment #6 from icegood <icegood1980 at gmail dot com> 2012-11-19 17:35:47 UTC ---
Created attachment 6744
  --> http://sourceware.org/bugzilla/attachment.cgi?id=6744
backtrace

-- 
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 c++/14834] Crash while trying to read field

[tromey at redhat dot com]

http://sourceware.org/bugzilla/show_bug.cgi?id=14834

--- Comment #7 from Tom Tromey <tromey at redhat dot com> 2012-11-19 17:44:41 UTC ---
Could you also post "ptype" of the problem variable?

-- 
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 c++/14834] Crash while trying to read field

[icegood1980 at gmail dot com]

http://sourceware.org/bugzilla/show_bug.cgi?id=14834

--- Comment #8 from icegood <icegood1980 at gmail dot com> 2012-11-19 18:15:19 UTC ---
Created attachment 6745
  --> http://sourceware.org/bugzilla/attachment.cgi?id=6745
backtrace2

Just compiled gdb with debug symbols. Backtrace recreated. Got rid of
<optimized out>. Hope it helps more. And sorry for windows-style EOL (my lovely
editor is under wine)

ptype:

(gdb) ptype this
type = class VolumeChangeMove<EnsembleNPT> : public McMove {
  private:
    double MaxRelativeLogChange;
    double Volume_Init;
    std::vector<_T_VolumeChangeMove_NeighboursByVolume,
std::allocator<_T_VolumeChangeMove_NeighboursByVolume> > *neighbours;
    CellList *FCellList;

  public:
    void VolumeChangeMove(McSystem *);
    ~VolumeChangeMove(int);
    virtual void read(FILE *);
    virtual void initialize(void);
    virtual bool move(void);
  private:
    void initialize_common(void);
    void CacheNeighbours(void);
    void SetNeighboursByVolume(double);
} * const
(gdb) ptype MaxRelativeLogChange
type = double

-- 
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 c++/14834] Crash while trying to read field

[tromey at redhat dot com]

http://sourceware.org/bugzilla/show_bug.cgi?id=14834

--- Comment #9 from Tom Tromey <tromey at redhat dot com> 2012-11-19 19:01:18 UTC ---
How about also ptype of all the base classes?
McMove, IoList, Io, and BaseClass?
I played around a bit trying to construct a failing case
but I still haven't been able to.

-- 
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 c++/14834] Crash while trying to read field

[icegood1980 at gmail dot com]

http://sourceware.org/bugzilla/show_bug.cgi?id=14834

--- Comment #10 from icegood <icegood1980 at gmail dot com> 2012-11-19 19:35:03 UTC ---
(gdb) ptype McMove
Type VolumeChangeMove<EnsembleNPT> has no component named McMove.
(gdb) p (McMove *) this
A syntax error in expression, near `) this'.

How to get context of McMove???

-- 
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 c++/14834] Crash while trying to read field

[tromey at redhat dot com]

http://sourceware.org/bugzilla/show_bug.cgi?id=14834

--- Comment #11 from Tom Tromey <tromey at redhat dot com> 2012-11-19 19:49:36 UTC ---
I guess you could try "ptype ::McMove"

-- 
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 c++/14834] Crash while trying to read field

[icegood1980 at gmail dot com]

http://sourceware.org/bugzilla/show_bug.cgi?id=14834

--- Comment #12 from icegood <icegood1980 at gmail dot com> 2012-11-19 20:13:44 UTC ---
Created attachment 6746
  --> http://sourceware.org/bugzilla/attachment.cgi?id=6746
ptypes

Meanwhile i'm trying to dig from other side. 
In findvar.c:176 have 
(gdb)  p type->main_type->code
$4 = TYPE_CODE_FLT

as it should be. The question why it treats to be pointer.

-- 
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 c++/14834] Crash while trying to read field

[tromey at redhat dot com]

http://sourceware.org/bugzilla/show_bug.cgi?id=14834

--- Comment #13 from Tom Tromey <tromey at redhat dot com> 2012-11-19 20:19:06 UTC ---
(In reply to comment #12)
> Created attachment 6746 [details]
> ptypes
> 
> Meanwhile i'm trying to dig from other side. 
> In findvar.c:176 have 
> (gdb)  p type->main_type->code
> $4 = TYPE_CODE_FLT
> 
> as it should be. The question why it treats to be pointer.

This code is trying to print the vtable pointer specially.
What is happening is that it is confused and thinks this field
is the vtable pointer.  What I'm trying to do is reconstruct
the failure here, hoping that with the right type hierarchy
I can recreate it locally.

-- 
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 c++/14834] Crash while trying to read field

[tromey at redhat dot com]

http://sourceware.org/bugzilla/show_bug.cgi?id=14834

--- Comment #14 from Tom Tromey <tromey at redhat dot com> 2012-11-19 20:35:56 UTC ---
Unfortunately I still can't make a test case.

The immediate problem here is in cp_print_value_fields.
We have i == vptr_fieldno && type == vptr_basetype
-- but erroneously.

So one question is the value of 'i' and also 'type'
(here, "type.main_type.name" will tell you the type currently
being printed).

But I suspect that this is just the local problem and the
real issue is that get_vptr_fieldno is returning the wrong
results.  And if that is so then it is likely that the issue
is either in the debuginfo or the debuginfo reader.

Any chance you can send me your executable?

-- 
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 c++/14834] Crash while trying to read field

[icegood1980 at gmail dot com]

http://sourceware.org/bugzilla/show_bug.cgi?id=14834

--- Comment #15 from icegood <icegood1980 at gmail dot com> 2012-11-19 21:30:34 UTC ---
Created attachment 6747
  --> http://sourceware.org/bugzilla/attachment.cgi?id=6747
exe

Additionaly if you need code:
svn checkout
svn://svn.code.sf.net/p/polysim/code/trunk/main/monte_carlo_offlattice_moldyn/ 
polysim-code

-- 
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 c++/14834] Crash while trying to read field

[tromey at redhat dot com]

http://sourceware.org/bugzilla/show_bug.cgi?id=14834

--- Comment #16 from Tom Tromey <tromey at redhat dot com> 2012-11-19 21:51:40 UTC ---
I tried setting a breakpoint in the constructor:

b VolumeChangeMove<EnsembleNPT>::VolumeChangeMove

and then running the program.
Then printing "*this" works fine for me, using today's gdb.

So, still mysterious :(

-- 
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 c++/14834] Crash while trying to read field

[icegood1980 at gmail dot com]

http://sourceware.org/bugzilla/show_bug.cgi?id=14834

icegood <icegood1980 at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|REOPENED                    |RESOLVED
         Resolution|                            |FIXED

--- Comment #17 from icegood <icegood1980 at gmail dot com> 2012-11-20 12:26:38 UTC ---
No more mysterious. 
On 7.5.50.20121119-cvs everything works!
Seems some updates was done during exactly last week.
In fact i didn't do "cvs update" so i tested exactly only week update...


========
and still cannot do "cvs update"

root@ubuntu:~/Documents/src# cvs -d :pserver:anoncvs@sourceware.org:/cvs/src
login
Logging in to :pserver:anoncvs@sourceware.org:2401/cvs/src
CVS password: 
cvs [login aborted]: unrecognized auth response from sourceware.org: Service
temporarily unavailable, please try again later

whatever...

-- 
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.