2

Looking for various tools (free/commercial) available for detection of memory leaks static/runtime on HP-UX Itanium platform.

Background, we: Use HP-UX 11.31 ia64. But, all our applications are still 32bits only. Have software with object files from C/Pro*C/COBOL and a very large application with lot of files/programs. C files are compiled with standard C compiler (cc), Pro*C with Oracle's proc and COBOL with Microfocus' cob. Finally, all the object files are linked with cob linker. Facing core dumps, due to memory leaks/invalid references (mostly from C/Pro*C code)

What was tried: Used gdb and RTC (HP RunTimeCheck for memory analysis), but due to mixed nature of COBOL and C, the tool is not able to give vital clues. Planned to use Insure++, but found that, it's not supported on HP-Itanium. Currently, relying on static debugging and manual prints, but as you can see, very slow and ineffective.

Can anybody please suggest tools/software available to do effective memory leaks detection in this scenario.

Thanks in advance.

ps: While searching on the web, I came across one commercial tool, but never used it though. http://www.dynamic-memory.com/products_Overview_htm.php

compuneo
  • 351
  • 1
  • 3
  • 13

2 Answers2

1

HP WDB is recognized by HP for these purposes: HP WDB

Tio Pepe
  • 3,071
  • 1
  • 17
  • 22
0

Our CheckPointer tool that finds memory management mistakes in C programs. If you have not made any such errors, on exit it will tell you where unfreed memory was allocated.

Because it operates on source code, it isn't specifically dependent on the Itanium hardware, but it is compiler-dependent (handles GCC 3/4 + Microsoft C dialects). The ProC you would handle by preprocessing the ProC code to produce C and then applying Checkpointer to the generated C code.

You will likely have to build some wrappers for your COBOL code (to verify that the COBOL code doesn't do something bad with a pointer). COBOL doesn't really do a lot of dynamic allocation/pointer dereferencing (watch out for CALL variable statements) so such wrapper models shouldn't be complicated.

Ira Baxter
  • 93,541
  • 22
  • 172
  • 341