On Apr 25, 2008, at 6:49 AM, Chakravarthula, Krishnamurthy wrote:
I get the following error when I build log4cxx 0.10.0. Please advice.
Did anybody get this working on HP Unix. If so, can you please pass on
the configurations and build steps.
objectptr.cpp: In static member function 'static void*
log4cxx::helpers::ObjectPtrBase::exchange(void**, void*)':
objectptr.cpp:41: error: cast from 'void*' to 'apr_uint32_t' loses
precisiongmake[3]: *** [objectptr.lo] Error 1
gmake[3]: Leaving directory
`/u/zztstkqa/jen/apache-log4cxx-0.10.0/src/main/cpp'
gmake[2]: *** [all-recursive] Error 1
gmake[2]: Leaving directory
`/u/zztstkqa/jen/apache-log4cxx-0.10.0/src/main'
gmake[1]: *** [all-recursive] Error 1
gmake[1]: Leaving directory `/u/zztstkqa/jen/apache-log4cxx-0.10.0/
src'
gmake: *** [all-recursive] Error 1
Glad you got past the APR build problems.
The code that is failing to compile is this little chunk which is
called by the ObjectPtrT<T> template during assignment operations to
exchange the new and old values. It is preferable, but not absolutely
essential, that this exchange be atomic, so #if's are used to check
for the presence of atomic exchange methods.
The problem is that the middle chunk in the following code fragment is
active, but doesn't compile. If the APR configuration has detected
that void* is 4 bytes in size, then apr_atomic_xchg32 is used. The
compilation error indicates that void* can't be cast to apr_uint32_t
which is a 32-bit unsigned integer which you'd expect would be able to
hold a void* if void* was 4 bytes in size.
void* ObjectPtrBase::exchange(void** destination, void* newValue) {
#if _WIN32 && (!defined(_MSC_VER) || _MSC_VER >= 1300)
return InterlockedExchangePointer(destination, newValue);
#elif APR_SIZEOF_VOIDP == 4
return (void*) apr_atomic_xchg32((volatile apr_uint32_t*)
destination,
(apr_uint32_t) newValue);
#else
void* oldValue = *destination;
*destination = newValue;
return oldValue;
#endif
You could work around the compile problem, by modifying the #elif so
that the last fragment is active. Either a hack
#elif 0
or a refinement of the condition (which please submit back to the list)
#elif APR_SIZEOF_VOIDP == 4 && !defined(SOMEFLAG)
However, the compile error is problematic since it suggests that APR
might have falsely detected a 32-bit void* when void* is 64-bit or APR
was compiled as a 32-bit app and log4cxx is being built as a 64-bit
app. In either case, bad things would be likely.
If you thought that you were building a 64-bit APR and
APR_SIZEOF_VOIDP == 4, then something is wrong in APR's configure on
your platform and you should raise the issue on the [EMAIL PROTECTED] mailing
list.
