Hi John,
What I have gone for is to convert ShapeAttribute::_string usage so
that it uses strdup and free rather than the previous awkward new
char[] copy code path. The new code stores a null pointer when a null
value is passed in. This could possible break other code if it
assumes that _string is never null, but if this is the case then this
code itself is broken and needs fixing.
I have now checked in this change, but don't have any problematic
files to test against. Could you send me one? Or do an svn checkout
and let me know how you get on.
Cheers,
Robert.
On Wed, Jun 18, 2008 at 2:42 PM, Robert Osfield
<[EMAIL PROTECTED]> wrote:
> Hi John,
>
> Reviewing the fix, it strikes me that the code should probably be
> something like:
>
> ShapeAttribute::ShapeAttribute(const char * name, const char * value) :
> _name(name),
> _type(STRING)
> {
> if (value) _string = strdup(value);
> else // some kind of fallback
> }
>
> A null name would also be potential problem for the ShapeAttribute code.
>
> Thoughts?
>
> Robert.
>
>
> On Thu, Jun 12, 2008 at 3:59 PM, John Vidar Larring
> <[EMAIL PROTECTED]> wrote:
>> Hi All,
>>
>> Short version:
>> --------------
>>
>> Problem: Empty string value in an osgSim::ShapeAttributeList causes SIGSEV
>> when loading a .osg file.
>>
>> Solution: Changing line 47 in src/osgSim/ShapeAttribute.cpp from:
>> - std::string str(value);
>> + std::string str((value ? value : ""));
>>
>> Attached file is a modification of src/osgSim/ShapeAttribute.cpp in svn
>> revision 8426.
>>
>> Long version (including context):
>> ---------------------------------
>>
>> osgGIS now allows the user to keep meta-data when converting ESRI Shape
>> files to OSG graph:
>>
>> osggis: world.shp -> world.osg:
>> [...snip...]
>> Geode {
>> UserData {
>> osgSim::ShapeAttributeList {
>> string "cntry_name" "Antarctica"
>> int "colormap" 7
>> string "curr_code" "NCIC"
>> string "curr_type" ""
>> string "fips_cntry" "AY"
>> string "gmi_cntry" "ATA"
>> string "iso_2digit" "AQ"
>> string "iso_3digit" "ATA"
>> string "landlocked" "N"
>> string "long_name" "Antarctica"
>> int "pop_cntry" -99999
>> string "sovereign" "Antarctica"
>> double "sqkm" 1.23031e+07
>> double "sqmi" 4.75021e+06
>> }
>> }
>> [...snip...]
>>
>> ...which is parallell to what 'osgconv world.shp world.osg' produces:
>> [...snip...]
>> Geometry {
>> DataVariance STATIC
>> UserData {
>> osgSim::ShapeAttributeList {
>> string "FIPS_CNTRY" "AY"
>> string "GMI_CNTRY" "ATA"
>> string "ISO_2DIGIT" "AQ"
>> string "ISO_3DIGIT" "ATA"
>> string "CNTRY_NAME" "Antarctica
>>
>> "
>> string "LONG_NAME" "Antarctica "
>> string "SOVEREIGN" "Antarctica "
>> int "POP_CNTRY" -99999
>> string "CURR_TYPE" " "
>> string "CURR_CODE" "NCIC"
>> string "LANDLOCKED" "N"
>> int "SQKM" 12303052
>> int "SQMI" 4750207
>> int "COLORMAP" 7
>> }
>> }
>> [...snip...]
>>
>> The difference is that osgconv (a.k.a. shp plugin) adds whitespace to string
>> "curr_type" while osggis does not. However, the output from osggis is
>> cleaner and should IMHO be acceptable.
>>
>> However, osgDB::Input in ShapeAttributeList_readLocalData converts the empty
>> string to a NULL pointer:
>>
>> bool ShapeAttributeList_readLocalData(osg::Object &obj, osgDB::Input &fr)
>> {
>> bool iteratorAdvanced = false;
>> ShapeAttributeList &sal = static_cast<ShapeAttributeList &>(obj);
>>
>> int entry = fr[0].getNoNestedBrackets();
>>
>> while (!fr.eof() && fr[0].getNoNestedBrackets()>=entry)
>> {
>> if (fr.matchSequence("string %s %s"))
>> {
>> sal.push_back(osgSim::ShapeAttribute(fr[1].getStr(),
>> fr[2].getStr())); <<== fr[2].getStr() returns NULL for 'string "curr_type"
>> ""'
>> fr += 3;
>> iteratorAdvanced = true;
>> }
>> [...snip...]
>>
>> This results in a SIGSEV in ShapeAttribute::ShapeAttribute(const char *
>> name, const char * value) with the following stack trace:
>>
>> #0 0x00000030a032e25d in raise () from /lib64/tls/libc.so.6
>> #1 0x00000030a032fa5e in abort () from /lib64/tls/libc.so.6
>> #2 0x00000030a28b1148 in __gnu_cxx::__verbose_terminate_handler () from
>> /usr/lib64/libstdc++.so.6
>> #3 0x00000030a28af176 in __cxa_call_unexpected () from
>> /usr/lib64/libstdc++.so.6
>> #4 0x00000030a28af1a3 in std::terminate () from /usr/lib64/libstdc++.so.6
>> #5 0x00000030a28af2a3 in __cxa_throw () from /usr/lib64/libstdc++.so.6
>> #6 0x00000030a2851bf0 in std::__throw_logic_error () from
>> /usr/lib64/libstdc++.so.6
>> #7 0x00000030a2892706 in std::basic_string<char, std::char_traits<char>,
>> std::allocator<char> >::basic_string$base () from /usr/lib64/libstdc++.so.6
>> #8 0x00000030a28927b3 in std::basic_string<char, std::char_traits<char>,
>> std::allocator<char> >::basic_string () from /usr/lib64/libstdc++.so.6
>> #9 0x0000002a96c9c2d7 in osgSim::ShapeAttribute::ShapeAttribute () at
>> /usr/lib/gcc/x86_64-redhat-linux/3.4.6/../../../../include/c++/3.4.6/bits/locale_facets.tcc:2443
>> #10 0x0000002a98083d16 in ShapeAttributeList_readLocalData ([EMAIL
>> PROTECTED],
>> [EMAIL PROTECTED]) at
>> /home/john/external.el4/OpenSceneGraph/trunk/src/osgPlugins/osgSim/IO_ShapeAttribute.cpp:38
>> #11 0x0000002a95cab5b0 in osgDB::Registry::readObject (this=0x51ac90,
>> [EMAIL PROTECTED], [EMAIL PROTECTED]) at
>> /home/john/external.el4/OpenSceneGraph/trunk/src/osgDB/Registry.cpp:1003
>> #12 0x0000002a95cab78a in osgDB::Registry::readObject (this=0x51ac90,
>> [EMAIL PROTECTED]) at
>> /home/john/external.el4/OpenSceneGraph/trunk/src/osgDB/Registry.cpp:1036
>>
>> The attached file fixes the problem by testing for NULL pointer value in
>> ShapeAttribute::ShapeAttribute.
>>
>> Best regards,
>> John
>>
>> --
>> This message has been scanned for viruses and
>> dangerous content by MailScanner, and is
>> believed to be clean.
>>
>>
>> _______________________________________________
>> osg-submissions mailing list
>> [email protected]
>> http://lists.openscenegraph.org/listinfo.cgi/osg-submissions-openscenegraph.org
>>
>>
>
_______________________________________________
osg-submissions mailing list
[email protected]
http://lists.openscenegraph.org/listinfo.cgi/osg-submissions-openscenegraph.org
