https://bugs.kde.org/show_bug.cgi?id=398445
Philippe Waroquiers <philippe.waroqui...@skynet.be> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |philippe.waroquiers@skynet.
| |be
--- Comment #1 from Philippe Waroquiers <philippe.waroqui...@skynet.be> ---
At my work, we are also using shared memory (shared between several processes).
In the same process, the shared memory is mapped only once.
In that setup, a piece of shared memory can be initialised by one process,
but read by another process.
So, for this, we are using a valgrind client request to declare the
shared memory as initialised after attach.
In your test case, the same memory is mapped twice (at different addresses)
by the same process.
This is an unusual usage pattern of shared memory.
Is that representative of your real application ?
If yes, then I think it will be quite unlikely this can be improved
in valgrind: even if valgrind could maybe track the system calls
to detect that the same piece of memory is in fact mapped twice,
handling this in memcheck will very probably imply a specific
data structure to be looked at for each access, and that will bring
a very high cost on programs not using this unusual shared memory usage.
So, in summary, it would be better to map the memory only once
and/or declare it initialised after map.
--
You are receiving this mail because:
You are watching all bug changes.
