[serialization][trunk] Latest patch for void_cast.cpp causes segfaults

[Hartmut Kaiser]

Hi all,

apparently the latest patch to boost/libs/serialization/src/void_cast.cpp
(rev.53078, svn:date: 2009-05-18T06:25:56.023190Z, svn:author: ramey,
svn:log: fix errors in extended_type_info management) causes an existing
application to segfault during cleanup. The application was working properly
before. The attached patch fixes the issue.

Ok to commit?

Regards Hartmut



_______________________________________________
Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost

[Hartmut Kaiser]

> apparently the latest patch to
> boost/libs/serialization/src/void_cast.cpp
> (rev.53078, svn:date: 2009-05-18T06:25:56.023190Z, svn:author: ramey,
> svn:log: fix errors in extended_type_info management) causes an
> existing application to segfault during cleanup. The application was
> working properly before. The attached patch fixes the issue.
>
> Ok to commit?

Ping?
Regards Hartmut



_______________________________________________
Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost

[Sohail Somani-2]

Hartmut Kaiser wrote:

>> apparently the latest patch to
>> boost/libs/serialization/src/void_cast.cpp
>> (rev.53078, svn:date: 2009-05-18T06:25:56.023190Z, svn:author: ramey,
>> svn:log: fix errors in extended_type_info management) causes an
>> existing application to segfault during cleanup. The application was
>> working properly before. The attached patch fixes the issue.
>>
>> Ok to commit?
>
> Ping?
> Regards Hartmut

I know it's not a "memory leak" the way you've done it because it's in
cleanup, but I changed the code to the following:

                 set_type::value_type p = *it;
                 s.erase(it);
                 delete p;
                 it = s.begin();

If I had delete *it before the erase() call, all hell broke loose and I
do not understand why (and have not had time to figure out yet.)

--
Sohail Somani
http://uint32t.blogspot.com

_______________________________________________
Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost

[Robert Ramey]

Sorry, I've been busy with some other stuff and a short trip.  I hope to
look at this in the next few days.

Robert Ramey

Hartmut Kaiser wrote:

>>> apparently the latest patch to
>>> boost/libs/serialization/src/void_cast.cpp
>>> (rev.53078, svn:date: 2009-05-18T06:25:56.023190Z, svn:author:
>>> ramey, svn:log: fix errors in extended_type_info management) causes
>>> an existing application to segfault during cleanup. The application
>>> was working properly before. The attached patch fixes the issue.
>>>
>>> Ok to commit?
>>
>> Ping?
>> Regards Hartmut
>>
>>
>
>
>
>> _______________________________________________
>> Unsubscribe & other changes:
>> http://lists.boost.org/mailman/listinfo.cgi/boost 

_______________________________________________
Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost

[Hartmut Kaiser]

> Sorry, I've been busy with some other stuff and a short trip.  I hope
> to
> look at this in the next few days.

Any news on that? It would be really unfortunate if this bug crept into the
upcoming release, making it unusable for us.

Regards Hartmut

>
> Robert Ramey
>
> Hartmut Kaiser wrote:
> >>> apparently the latest patch to
> >>> boost/libs/serialization/src/void_cast.cpp
> >>> (rev.53078, svn:date: 2009-05-18T06:25:56.023190Z, svn:author:
> >>> ramey, svn:log: fix errors in extended_type_info management) causes
> >>> an existing application to segfault during cleanup. The application
> >>> was working properly before. The attached patch fixes the issue.
> >>>
> >>> Ok to commit?
> >>
> >> Ping?
> >> Regards Hartmut
> >>
> >>
> >
> >
> >
> >> _______________________________________________
> >> Unsubscribe & other changes:
> >> http://lists.boost.org/mailman/listinfo.cgi/boost
>
>
>
> _______________________________________________
> Unsubscribe & other changes:
> http://lists.boost.org/mailman/listinfo.cgi/boost

_______________________________________________
Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost

[Robert Ramey]

I've got it on my desk top along with a couple others. I'm hoping to get all
of these fixes in the next few days. I realize I'm overdue, but I'm
stretched thin right now.

Robert Ramey

Hartmut Kaiser wrote:

>> Sorry, I've been busy with some other stuff and a short trip.  I hope
>> to
>> look at this in the next few days.
>
> Any news on that? It would be really unfortunate if this bug crept
> into the upcoming release, making it unusable for us.
>
> Regards Hartmut
>
>>
>> Robert Ramey
>>
>> Hartmut Kaiser wrote:
>>>>> apparently the latest patch to
>>>>> boost/libs/serialization/src/void_cast.cpp
>>>>> (rev.53078, svn:date: 2009-05-18T06:25:56.023190Z, svn:author:
>>>>> ramey, svn:log: fix errors in extended_type_info management)
>>>>> causes an existing application to segfault during cleanup. The
>>>>> application was working properly before. The attached patch fixes
>>>>> the issue.
>>>>>
>>>>> Ok to commit?
>>>>
>>>> Ping?
>>>> Regards Hartmut
>>>>
>>>>
>>>
>>>
>>>
>>>> _______________________________________________
>>>> Unsubscribe & other changes:
>>>> http://lists.boost.org/mailman/listinfo.cgi/boost
>>
>>
>>
>> _______________________________________________
>> Unsubscribe & other changes:
>> http://lists.boost.org/mailman/listinfo.cgi/boost
>
> _______________________________________________
> Unsubscribe & other changes:
> http://lists.boost.org/mailman/listinfo.cgi/boost 

_______________________________________________
Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost

[Hartmut Kaiser]

Robert,

> I've got it on my desk top along with a couple others. I'm hoping to
> get all
> of these fixes in the next few days. I realize I'm overdue, but I'm
> stretched thin right now.

Thanks for your fixes in SVN. I don't know whether those were supposed to
fix this particular problem, but I'm afraid needing to report that it's not
fixed yet. I'm still seeing the same segmentation fault at application
shutdown :-(

Thanks!
Regards Hartmut

>
> Robert Ramey
>
> Hartmut Kaiser wrote:
> >> Sorry, I've been busy with some other stuff and a short trip.  I
> hope
> >> to
> >> look at this in the next few days.
> >
> > Any news on that? It would be really unfortunate if this bug crept
> > into the upcoming release, making it unusable for us.
> >
> > Regards Hartmut
> >
> >>
> >> Robert Ramey
> >>
> >> Hartmut Kaiser wrote:
> >>>>> apparently the latest patch to
> >>>>> boost/libs/serialization/src/void_cast.cpp
> >>>>> (rev.53078, svn:date: 2009-05-18T06:25:56.023190Z, svn:author:
> >>>>> ramey, svn:log: fix errors in extended_type_info management)
> >>>>> causes an existing application to segfault during cleanup. The
> >>>>> application was working properly before. The attached patch fixes
> >>>>> the issue.
> >>>>>
> >>>>> Ok to commit?
> >>>>
> >>>> Ping?
> >>>> Regards Hartmut
> >>>>
> >>>>
> >>>
> >>>
> >>>
> >>>> _______________________________________________
> >>>> Unsubscribe & other changes:
> >>>> http://lists.boost.org/mailman/listinfo.cgi/boost
> >>
> >>
> >>
> >> _______________________________________________
> >> Unsubscribe & other changes:
> >> http://lists.boost.org/mailman/listinfo.cgi/boost
> >
> > _______________________________________________
> > Unsubscribe & other changes:
> > http://lists.boost.org/mailman/listinfo.cgi/boost
>
>
>
> _______________________________________________
> Unsubscribe & other changes:
> http://lists.boost.org/mailman/listinfo.cgi/boost

_______________________________________________
Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost

[Robert Ramey]

I pretty sure I addressed this where it was first appearing.  All tests in
test suite
which were faililng before are now passing - EXCEPT one - test_no_rtti.
It just happens that this test trips upon the same problem in a different
place.  I'm looking into this now and hope to have it resolved soon.

Robert Ramey

Hartmut Kaiser wrote:

> Robert,
>
>> I've got it on my desk top along with a couple others. I'm hoping to
>> get all
>> of these fixes in the next few days. I realize I'm overdue, but I'm
>> stretched thin right now.
>
> Thanks for your fixes in SVN. I don't know whether those were
> supposed to
> fix this particular problem, but I'm afraid needing to report that
> it's not fixed yet. I'm still seeing the same segmentation fault at
> application shutdown :-(
>
> Thanks!
> Regards Hartmut
>
>>
>> Robert Ramey
>>
>> Hartmut Kaiser wrote:
>>>> Sorry, I've been busy with some other stuff and a short trip.  I
>> hope
>>>> to
>>>> look at this in the next few days.
>>>
>>> Any news on that? It would be really unfortunate if this bug crept
>>> into the upcoming release, making it unusable for us.
>>>
>>> Regards Hartmut
>>>
>>>>
>>>> Robert Ramey
>>>>
>>>> Hartmut Kaiser wrote:
>>>>>>> apparently the latest patch to
>>>>>>> boost/libs/serialization/src/void_cast.cpp
>>>>>>> (rev.53078, svn:date: 2009-05-18T06:25:56.023190Z, svn:author:
>>>>>>> ramey, svn:log: fix errors in extended_type_info management)
>>>>>>> causes an existing application to segfault during cleanup. The
>>>>>>> application was working properly before. The attached patch
>>>>>>> fixes the issue.
>>>>>>>
>>>>>>> Ok to commit?
>>>>>>
>>>>>> Ping?
>>>>>> Regards Hartmut
>>>>>>
>>>>>>
>>>>>
>>>>>
>>>>>
>>>>>> _______________________________________________
>>>>>> Unsubscribe & other changes:
>>>>>> http://lists.boost.org/mailman/listinfo.cgi/boost
>>>>
>>>>
>>>>
>>>> _______________________________________________
>>>> Unsubscribe & other changes:
>>>> http://lists.boost.org/mailman/listinfo.cgi/boost
>>>
>>> _______________________________________________
>>> Unsubscribe & other changes:
>>> http://lists.boost.org/mailman/listinfo.cgi/boost
>>
>>
>>
>> _______________________________________________
>> Unsubscribe & other changes:
>> http://lists.boost.org/mailman/listinfo.cgi/boost
>
> _______________________________________________
> Unsubscribe & other changes:
> http://lists.boost.org/mailman/listinfo.cgi/boost 

_______________________________________________
Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost

[Hartmut Kaiser]

> I pretty sure I addressed this where it was first appearing.  All tests
> in
> test suite
> which were faililng before are now passing - EXCEPT one - test_no_rtti.
> It just happens that this test trips upon the same problem in a
> different
> place.  I'm looking into this now and hope to have it resolved soon.

I'm attaching a patch fixing the problem for me, but I'm not sure if this
isn't introducing a memory leak. I still think the delete is bogus. Doesn't
the erase(*it) already delete the container element?

Regards Hartmut

>
> Robert Ramey
>
> Hartmut Kaiser wrote:
> > Robert,
> >
> >> I've got it on my desk top along with a couple others. I'm hoping to
> >> get all
> >> of these fixes in the next few days. I realize I'm overdue, but I'm
> >> stretched thin right now.
> >
> > Thanks for your fixes in SVN. I don't know whether those were
> > supposed to
> > fix this particular problem, but I'm afraid needing to report that
> > it's not fixed yet. I'm still seeing the same segmentation fault at
> > application shutdown :-(
> >
> > Thanks!
> > Regards Hartmut
> >
> >>
> >> Robert Ramey
> >>
> >> Hartmut Kaiser wrote:
> >>>> Sorry, I've been busy with some other stuff and a short trip.  I
> >> hope
> >>>> to
> >>>> look at this in the next few days.
> >>>
> >>> Any news on that? It would be really unfortunate if this bug crept
> >>> into the upcoming release, making it unusable for us.
> >>>
> >>> Regards Hartmut
> >>>
> >>>>
> >>>> Robert Ramey
> >>>>
> >>>> Hartmut Kaiser wrote:
> >>>>>>> apparently the latest patch to
> >>>>>>> boost/libs/serialization/src/void_cast.cpp
> >>>>>>> (rev.53078, svn:date: 2009-05-18T06:25:56.023190Z, svn:author:
> >>>>>>> ramey, svn:log: fix errors in extended_type_info management)
> >>>>>>> causes an existing application to segfault during cleanup. The
> >>>>>>> application was working properly before. The attached patch
> >>>>>>> fixes the issue.
> >>>>>>>
> >>>>>>> Ok to commit?
> >>>>>>
> >>>>>> Ping?
> >>>>>> Regards Hartmut
> >>>>>>
> >>>>>>
> >>>>>
> >>>>>
> >>>>>
> >>>>>> _______________________________________________
> >>>>>> Unsubscribe & other changes:
> >>>>>> http://lists.boost.org/mailman/listinfo.cgi/boost
> >>>>
> >>>>
> >>>>
> >>>> _______________________________________________
> >>>> Unsubscribe & other changes:
> >>>> http://lists.boost.org/mailman/listinfo.cgi/boost
> >>>
> >>> _______________________________________________
> >>> Unsubscribe & other changes:
> >>> http://lists.boost.org/mailman/listinfo.cgi/boost
> >>
> >>
> >>
> >> _______________________________________________
> >> Unsubscribe & other changes:
> >> http://lists.boost.org/mailman/listinfo.cgi/boost
> >
> > _______________________________________________
> > Unsubscribe & other changes:
> > http://lists.boost.org/mailman/listinfo.cgi/boost
>
>
>
> _______________________________________________
> Unsubscribe & other changes:
> http://lists.boost.org/mailman/listinfo.cgi/boost

_______________________________________________
Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost

[Robert Ramey]

If you're patch fixes it for you, It's still not right so I'll take another
look at it. I think your solution does in fact create a memory leak
which it is tempting to ignore.  Except that I'm concerned about
the dynamic unloading of DLLS which will need to remove
these records for the types no longer supported without that DLL.

Looking at it, I think I do see something that can be addressed.

Robert Ramey

Hartmut Kaiser wrote:

>>> I pretty sure I addressed this where it was first appearing.  All
>>> tests in
>>> test suite
>>> which were faililng before are now passing - EXCEPT one -
>>> test_no_rtti. It just happens that this test trips upon the same
>>> problem in a different
>>> place.  I'm looking into this now and hope to have it resolved soon.
>>
>> I'm attaching a patch fixing the problem for me, but I'm not sure if
>> this isn't introducing a memory leak. I still think the delete is
>> bogus. Doesn't the erase(*it) already delete the container element?
>>
>> Regards Hartmut
>>
>>>
>>> Robert Ramey
>>>
>>> Hartmut Kaiser wrote:
>>>> Robert,
>>>>
>>>>> I've got it on my desk top along with a couple others. I'm hoping
>>>>> to get all
>>>>> of these fixes in the next few days. I realize I'm overdue, but
>>>>> I'm stretched thin right now.
>>>>
>>>> Thanks for your fixes in SVN. I don't know whether those were
>>>> supposed to
>>>> fix this particular problem, but I'm afraid needing to report that
>>>> it's not fixed yet. I'm still seeing the same segmentation fault at
>>>> application shutdown :-(
>>>>
>>>> Thanks!
>>>> Regards Hartmut
>>>>
>>>>>
>>>>> Robert Ramey
>>>>>
>>>>> Hartmut Kaiser wrote:
>>>>>>> Sorry, I've been busy with some other stuff and a short trip.  I
>>>>> hope
>>>>>>> to
>>>>>>> look at this in the next few days.
>>>>>>
>>>>>> Any news on that? It would be really unfortunate if this bug
>>>>>> crept into the upcoming release, making it unusable for us.
>>>>>>
>>>>>> Regards Hartmut
>>>>>>
>>>>>>>
>>>>>>> Robert Ramey
>>>>>>>
>>>>>>> Hartmut Kaiser wrote:
>>>>>>>>>> apparently the latest patch to
>>>>>>>>>> boost/libs/serialization/src/void_cast.cpp
>>>>>>>>>> (rev.53078, svn:date: 2009-05-18T06:25:56.023190Z,
>>>>>>>>>> svn:author: ramey, svn:log: fix errors in extended_type_info
>>>>>>>>>> management) causes an existing application to segfault
>>>>>>>>>> during cleanup. The application was working properly before.
>>>>>>>>>> The attached patch fixes the issue.
>>>>>>>>>>
>>>>>>>>>> Ok to commit?
>>>>>>>>>
>>>>>>>>> Ping?
>>>>>>>>> Regards Hartmut
>>>>>>>>>
>>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>> _______________________________________________
>>>>>>>>> Unsubscribe & other changes:
>>>>>>>>> http://lists.boost.org/mailman/listinfo.cgi/boost
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> _______________________________________________
>>>>>>> Unsubscribe & other changes:
>>>>>>> http://lists.boost.org/mailman/listinfo.cgi/boost
>>>>>>
>>>>>> _______________________________________________
>>>>>> Unsubscribe & other changes:
>>>>>> http://lists.boost.org/mailman/listinfo.cgi/boost
>>>>>
>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> Unsubscribe & other changes:
>>>>> http://lists.boost.org/mailman/listinfo.cgi/boost
>>>>
>>>> _______________________________________________
>>>> Unsubscribe & other changes:
>>>> http://lists.boost.org/mailman/listinfo.cgi/boost
>>>
>>>
>>>
>>> _______________________________________________
>>> Unsubscribe & other changes:
>>> http://lists.boost.org/mailman/listinfo.cgi/boost
>>
>
>
>
>> _______________________________________________
>> Unsubscribe & other changes:
>> http://lists.boost.org/mailman/listinfo.cgi/boost 

_______________________________________________
Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost