You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
xref issue #557 my new scheme for walking callstacks handles most fpo, but I'm avoiding any
expensive symbol lookups, and the debugger will likely do a better job in
some corner cases where unwind info is needed. this issue covers adding an
option to use dbghelp!StackWalk64 or other third-party stack walking
routines. if implemented there should be two options: one to use for
non-leak error reports, and another on leaks, since callstack walking on
every malloc for leak reports is much more performance-critical.
note that I briefly tried using dbghelp!StackWalk64 and found that it did
not seem to perform lazy symbol loading: only if I explicitly loaded
symbols up front in SymInitialize(), and examples of its usage sometimes
walk the whole module list themselves calling SymLoadModule64 explicitly.
in concert with drsyms I never got it working properly but I did not spend
much time on it.
there's also RtlCaptureStackBackTrace but I suspect it does not look at
symbol info
From bruen...@google.com on August 22, 2011 10:11:39
xref issue #557 my new scheme for walking callstacks handles most fpo, but I'm avoiding any
expensive symbol lookups, and the debugger will likely do a better job in
some corner cases where unwind info is needed. this issue covers adding an
option to use dbghelp!StackWalk64 or other third-party stack walking
routines. if implemented there should be two options: one to use for
non-leak error reports, and another on leaks, since callstack walking on
every malloc for leak reports is much more performance-critical.
note that I briefly tried using dbghelp!StackWalk64 and found that it did
not seem to perform lazy symbol loading: only if I explicitly loaded
symbols up front in SymInitialize(), and examples of its usage sometimes
walk the whole module list themselves calling SymLoadModule64 explicitly.
in concert with drsyms I never got it working properly but I did not spend
much time on it.
there's also RtlCaptureStackBackTrace but I suspect it does not look at
symbol info
there's also IDiaStackWalker but it uses COM
Original issue: http://code.google.com/p/drmemory/issues/detail?id=563
The text was updated successfully, but these errors were encountered: