PROBLEM: (79556) (PATCH ID: OSF445-001) ******** The output string for the swprintf() interface was not NULL terminated. PROBLEM: (79876) (PATCH ID: OSF445-023) ******** If the "where" or "t" ladebug command does not show all the currently active functions, it may be because of this problem. The program being debugged had to have been built with "split procedures", which is done via options in the om and spike tools. For spike, split procedures are generated if profiling is specified. For om, split procedures are generated if profiling AND the -split_procedure option are specified. PROBLEM: (TKTR30011, 78909) (PATCH ID: OSF445-025) ******** If a code segment contained a gmtime() call between calls to tzset(), or any function specified to call tzset() (such as localtime(), mktime(), strftime(), etc) , the tzname[0] array could contain erroneous time zone data referring to the "GMT" zone abbreviation instead of the local time zone, as specified in the TZ environment variable or default time zone for the system. PROBLEM: (62211) (PATCH ID: OSF445-031) ******** This fix increases the length of the user names for rsh and rexec to allow for NT interoperabilty. PROBLEM: (72100, 74961, 75971, 77080, 77288, 79588) (PATCH ID: OSF445-112) ******** Threaded applications which make extensive use of synchronization objects (mutexes, condition variables) may encounter performance and scaling regressions when run on some EV6 SMP machines. Customers encountering these regressions may observe negative scaling, decreased application throughput and increased CPU utilitzation, as the number of EV6 CPUs is increased. This version of DECthreads contains changes to the implementation of synchronization objects and thread scheduling around those objects to improve performance on all SMP systems. No functional or interface changes were made. PROBLEM: (BCSM807S5) (PATCH ID: OSF445-108) ******** Some applications assume that initial allocations of memory from the C run-time library's malloc() function will return zero-filled memory. The problem is that malloc() and free() often write into the first 16 bytes of the allocated memory. This change to malloc() zeros out the first 16 bytes in the allocated memory. The result is that memory allocated prior to any call to free() -- and prior to any other run-time library function that might call free -- should be zero filled.