When instrumentation mode is switched off, we want to fully
reset callgrinds internal execution state, which includes
the shadow call stack and function stack for context generation.
The latter was not completely reset before.
The bug showed up when switching the instrumentation mode on for
the second time, leading to
Callgrind: callstack.c:211 (vgCallgrind_push_call_stack):
Assertion 'current_entry->cxt != 0' failed.
This commit also improves debug output a little bit.
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@6045
Bug description: Very similar to cachegrind, callgrind stores
metainformation per guest instruction; this meta information is
given when calling into the simulator. In contrast to cachegrind,
callgrind keeps this info when the source is discarded, and checks
on retranslation whether the same meta info is generated.
This check sometimes fails: E.g. for rep x86 instructions, 2 simulator
calls
are usually generated for one x86 instruction (the instruction fetch and
a
data access), thus overwriting the data_size meta information for one
x86
instruction first with 0, and afterwards e.g. with 1. The check on
retranslation
fails because of this. The fix is to only write/check data_size values
>0.
git-svn-id: svn://svn.valgrind.org/valgrind/trunk@5942
