doxygen-develop

[Doxygen-develop] 1.3.1 bug with alphabetical index

[Mark T. <m...@id...>]

I've tried the new 1.3.1 that seems to have come out earlier in the
day.  Both the source and the binary die with a seg fault:

Generating alphabetical compound index...
QGArray::at: Absolute index 0 out of range

Both 1.3 and 1.2.13.1 that I have tried operated fine.  The
alphabetical index for the compounds has about 200 identifiers in it
so it should definitely be larger than 0. 

Let me know if there is anything I can do to help narrow it down. 

Here's the stack trace.

Generating alphabetical compound index...
QGArray::at: Absolute index 0 out of range

Program received signal SIGSEGV, Segmentation fault.
0x8084876 in writeAlphabeticalClassList (ol=@0x9e60460) at index.cpp:1301
1301          if (toupper(cd->className().at(index))!=startLetter) // new begin letter => new header
(gdb) where
#0  0x8084876 in writeAlphabeticalClassList (ol=@0x9e60460) at index.cpp:1301
#1  0x8085557 in writeAlphabeticalIndex (ol=@0x9e60460) at index.cpp:1456
#2  0x806009a in generateClassDocs () at doxygen.cpp:5694
#3  0x8069bb5 in generateOutput () at doxygen.cpp:8043
#4  0x8049b98 in main (argc=2, argv=0xbffff2e4) at main.cpp:37
#5  0x400a12eb in __libc_start_main () from /lib/libc.so.6

For what it's worth the offending bit is   

  ClassSDict::Iterator cli(Doxygen::classSDict); 

which is iterated a few times in the writeAlphabeticalClassList
function.  Apparently with my input the last item as a null QCString.
If I just if(cd->className().length()>0) all of the inner portions
where at(0) is called everything works fine.  So somewhere previously
the (null) class was being appended to the list.

Mark

Re: [Doxygen-develop] 1.3.1 bug with alphabetical index

[Dimitri v. H. <di...@st...>]

On Wed, May 28, 2003 at 04:40:16PM -0700, Mark Tigges wrote:
> 
> I've tried the new 1.3.1 that seems to have come out earlier in the
> day.  Both the source and the binary die with a seg fault:
> 
> Generating alphabetical compound index...
> QGArray::at: Absolute index 0 out of range
> 
> Both 1.3 and 1.2.13.1 that I have tried operated fine.  The
> alphabetical index for the compounds has about 200 identifiers in it
> so it should definitely be larger than 0. 

I have had another report about this. It appears this happens if
a class and a namespace at the global scope have the same name.
As a quick workaround you could try the add the following in the
contructor of ClassDef (in classdef.cpp around line 95)

  m_isAbstract = FALSE;
  m_isStatic = FALSE;
  m_membersMerged = FALSE;
  QCString ns;
  extractNamespaceName(m_name,m_className,ns);
+ if (m_className.isEmpty()) // correct for case where there is a namespace and
+                            // class with the same name
+ {
+   m_className = m_name.copy();
+   ns.resize(0);
+   
+ }

Let me know if this helps in your case.

Regards,
  Dimitri