1. "new"
On my machine
// Original
void* XMLPlatformUtils::makeMutex()
{
pthread_mutex_t* mutex = new pthread_mutex_t;
pthread_mutexattr_t* attr = new
pthread_mutexattr_t;
pthread_mutexattr_init(attr);
pthread_mutexattr_settype(attr,
PTHREAD_MUTEX_RECURSIVE);
if (pthread_mutex_init(mutex, attr))
{
ThrowXML(XMLPlatformUtilsException,
XML4CExcepts::Mutex_CouldNotCreate);
}
pthread_mutexattr_destroy(attr);
delete attr;
return (void*)(mutex);
}
takes twice as long as does
// Changed
void* XMLPlatformUtils::makeMutex()
{
pthread_mutex_t *mutex = new pthread_mutex_t;
pthread_mutexattr_t attr;
pthread_mutexattr_init(&attr);
pthread_mutexattr_settype(&attr,
PTHREAD_MUTEX_RECURSIVE);
if (pthread_mutex_init(mutex, &attr))
{
ThrowXML(XMLPlatformUtilsException,
XML4CExcepts::Mutex_CouldNotCreate);
}
pthread_mutexattr_destroy(&attr);
return (void *)(mutex);
}
/src/ contains 107 files in which new is being
called.
I've seen various places where local variables are
put on the heap, so maybe there is opportunity for
improvement.
If you think this would be worth doing (I have no
profiling data), I'd volunteer...
2. Would it be ok to use the STL if I were to change the source
code? I.e. what is
your strategy regarding C++ backward compatibility?
3. There is one problem. I built libxerces-c1_0.so with these options:
-p irix -c cc -x CC -m inmem -n fileonly -t native
Parsing my test-project results in
Fatal Error in file "/usr/people/marc/projects/xml/EPKO.xml", line 1,
column 48
Message: An exception occured! Type:RuntimeException,
Message:Could not create a
converter for encoding: iso-8859-1
What's amiss?
Thanks
-- marc --------------------------------------------------------------- marc stuessel mailto:[EMAIL PROTECTED] IST GmbH P.O. Box 11 12 29 Fax: ++49 (0) 721 / 831 32 33 76062 Karlsruhe Phone: ++49 (0) 721 / 831 32-0 Germany GSM: ++49 (0) 177 / 831 32 10
