Hi all,
currently, running json-integration-test with valgrind leads to the
following warning: "Syscall param write(buf) points to uninitialised
byte(s)". This is caused by PrimitiveBufferBuilder not initializing its
data memory. Note: we initialize null_bitmap_data_ by zeroing, i.e. setting
all v
Hi,
Writing uninitialized memory risks leaking private data. This might
lead to security issues.
I'd go for option 2. Option 3 sounds much more costly (we would be
zero-initializing large memory areas instead of small padding areas).
Regards
Antoine.
Le 03/07/2018 à 13:11, Dimitri Vorona
Hi,
I misunderstood the issue: the zeroing of the padding is correctly handled
by the writers, as long as they are careful not to read beyond the
specified size of the buffers (as opposed to its capacity).
The valgrind warning came from the way AppendNull(s) is implemented in some
of the builders
hi Dimitri,
Zeroing the padding is probably a good idea; I'd be interested to look
at the diff to see how many code paths are impacted. We already have a
number of places where we are zeroing buffers but as you have found it
is not 100% consistent.
- Wes
On Tue, Jul 3, 2018 at 10:13 AM, Dimitri
Joseph Toth created ARROW-2787:
--
Summary: Memory Issue passing table from python to c++ via cython
Key: ARROW-2787
URL: https://issues.apache.org/jira/browse/ARROW-2787
Project: Apache Arrow
Iss
Philipp Moritz created ARROW-2788:
-
Summary: [Plasma] Defining Delete semantics
Key: ARROW-2788
URL: https://issues.apache.org/jira/browse/ARROW-2788
Project: Apache Arrow
Issue Type: Improve
Brian Hulette created ARROW-2789:
Summary: [JS] Minor DataFrame improvements
Key: ARROW-2789
URL: https://issues.apache.org/jira/browse/ARROW-2789
Project: Apache Arrow
Issue Type: Improvemen