I ended up writing a short Bourne shell script that does the best I
know how to do when looking at just the binaries in the program/
directory. While there is no guarantee that the reported symbols are
completely unused, they are clearly canidates for the *_DLLPRIVATE
macros. Once made private, a per-module lint-like analysis could
finish determining whether a symbol is truly unused.
The script outputs two files. The first file contains the symbols that
are defined but not used by any UNDEF nor used in any relocation table.
The second file contains symbols that are defined but are used only
within a relocation table (i.e., only the defining binary uses it).
These results might be useful towards run-time linking optimization, as
the first file reports about 15,000 symbols and the second file reports
about 47,000 symbols that could be made private. Given that there are
over 100,000 symbols exported in all the binaries, this means there is
a potential for reducing the number of exported symbols by several
percent. I hesitate to say the full 15,000 could be privatized,
because many of the unused symbols are constructors, for example, which
probably have an intended purpose.
What is the best way to submit this script to OO.org for second
opinions and to see whether it would be useful towards run-time
optimization?
__________________________________
Do you Yahoo!?
Yahoo! Small Business - Try our new resources site!
http://smallbusiness.yahoo.com/resources/
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
