gprof total run time incorrect

[Paolo Carlini]

Hi all,

in these days I' experiencing a strange behaviour of gprof (2.11.2) on
my system (Linux2.4.4., gcc3.0, glibc2.2.3 (built with gcc2.95.3+atexit
patch)).

When profiling C++ programs (built with gcc3.0) the total running time
reported by gprof is incorrect by an apparently random factor of ~3-10
with respect to the total time (u+s) output by time. Otherwise, gprof
output seems reasonable.

Is anyone else seeing this?

Thanks,
Paolo Carlini.