Changelog for gprof: entries are in ascending order gprof.c:: ++jrb add reloc before dumping the summary file to the symbol value add _malloczero(1) :: the code seems to assume it. ------------------------------- Patchlevel 2 --------------------------- gprof.c, st-out.h:: andreas handle long GST format symbols ++jrb:: use of gcc -G to get long symbols is *HIGHLY* recommended (in fact i am considering making this the default. BTW thanks to apratt@atari, db now supports GST symbols. So does Michal Jaegermann's version of szadb) ------------------------------- Patchlevel 3 --------------------------- *.[ch]:: ++jrb -- sync code up with GNU gprof from binutils 1.9. -- add in g++ name demangler hooks. to use the name demangler all you have to do is delete the function cplus_demangle() from the end of gprof.c and link in cplus-dem.o instead. for now we leave this out because even with -G the mangled names are too long, and the demangler gets confused. gprof.c:: ++jrb filter out "gcc_compiled" symbols filter out the dummy symbols ld dumped in (in write_atari_sym) (ie. syms of type A_UNDF). Some of these are legit and not dummies, but certainly are of no interest to profiling. filter out symbols that look like "D:" where D is any drive. (this is to catch cases where there is no '\' in the pathname) ------------------------------- Patchlevel 4 --------------------------- gprof.c:: ++jrb pass underscores to the demangler. link in demangler cplusdem.c:: ++jrb c++ demangler Makefile:: ++jrb link in demangler ------------------------------- Patchlevel 5 --------------------------- gprof.c:: ++jrb various fixes from the net. gprof -s should work better now. ------------------------------- Patchlevel 6 --------------------------- cplusdem.c:: ++jrb fix up demangler as per util diffs ------------------------------- Patchlevel 7 --------------------------- cplusdem.c:: ++jrb more fixes to demangler as per util diffs gprof.c:: ++jrb dont strip off the initial '_' when reading symbols. this was preventing the demangler from doing its job properly. before printing the initial '_' if present is stripped, so this is not a problem when you have a unmangled name. ------------------------------- Patchlevel 8 --------------------------- gprof.c:: ++jrb clean up binary seach conditions in val_to_sym. ------------------------------- Patchlevel 9 ---------------------------