IBM AIX Problems Fixed
The following problems have been fixed. More details about each item follow this list. Items new with this release are in bold.
2 GB code size problem fixed
The error debugging code with data size greater than 2 GB is resolved. When this problem occurred, TotalView displayed the following error message:
TotalView: FATAL ERROR STARTING UP: Attempt to register overlapping
asects ...
(4646: 6.2.0-0)
Cray pointers issue
Problem diving on cray pointers in common blocks when using xlf 7.1 was resolved (4521: 6.1.0-2)
Diving on a Fortran 90 local variable problem fixed
Diving on a local variable in a Fortran 90 module would sometimes fail, causing a 'not found' dialog box to pop up. This occurred when the Prefered Scope field showed incorrect information regarding the current scope. (4617: 6.1.0-3)
Form_long result too large problem fixed
A problem with the C++ demangler would sometimes result in variable names that were too long to fit into an internal buffer. (4601: 6.2.0-0)
Fortran logical type issue
TotalView now displays Fortran logical types correctly (4708: 6.2.0-2)
Fortran parameter lookup problem fixed
The TotalView internal error resulting from attempting to lookup a Fortran parameter has been resolved. (4632: 6.2.0-0)
GCC 3.2: Internal Error starting up upon a GCC 3.2 compiled executable
A TotalView crash starting up upon an executable compiled with GCC 3.2 was fixed. The GCC 3.2 compiler failed to emit correct debug information, which caused an internal error within TotalView. (4507: 6.1.0-2)
Large data segments no longer crash TotalView
The error messages you saw were (in part):
TotalView: FATAL ERROR STARTING UP: read_all_symbols:
After save-and-stop, the target still isn't stopped
and
TotalView: FATAL ERROR STARTING UP: Attempt to register
overlapping asects (4524, 4646: 6.1.0-3)
MPI processes randomly stop before hitting a user breakpoint during startup
MPI processes randomly stopped before hitting a user breakpoint during startup of an MPI job. We are aware that this may occur under some circumstances and are working on an improved fix. (4634: 6.1.0-3)
Multi-dimensioned const arrays no longer crashes TotalView
This occurred when you examined these kinds arrays in C programs compiled with gcc 2.95 (4544: 6.1.03)
Multi-threaded core debugging problem fixed
TotalView no longer crashes with the following message when it is debugging a multi-threaded core file: (4807: 6.2.0-3)
Fatal error: While trying to workaround a GNU V3 compiler bug,
we failed to find a subroutine at provisional address ...
