Hello Adar Dembo,
I'd like you to do a code review. Please visit
http://gerrit.cloudera.org:8080/5734
to review the following change.
Change subject: binary_plain_block: fix an incorrect vector.reserve()
......................................................................
binary_plain_block: fix an incorrect vector.reserve()
I noticed the following surprising stack trace on a tserver:
Thread 118 (Thread 0x7f9a865ff700 (LWP 1332)):
#0 0x0000003ae0a89b40 in _wordcopy_fwd_aligned () from /lib64/libc.so.6
#1 0x0000003ae0a83992 in memmove () from /lib64/libc.so.6
#2 0x0000000001018dee in void std::vector<unsigned int,
std::allocator<unsigned int> >::_M_emplace_back_aux<unsigned int>(unsigned
int&&) ()
#3 0x00000000018c8b15 in emplace_back<unsigned int> (this=0x7f9a858304b0) at
/opt/rh/devtoolset-3/root/usr/include/c++/4.9.2/bits/vector.tcc:101
#4 push_back (this=0x7f9a858304b0) at
/opt/rh/devtoolset-3/root/usr/include/c++/4.9.2/bits/stl_vector.h:932
#5 kudu::cfile::BinaryPlainBlockDecoder::ParseHeader (this=0x7f9a858304b0) at
../../src/kudu/cfile/binary_plain_block.cc:225
#6 0x00000000018d4ffa in kudu::cfile::CFileIterator::PrepareForNewSeek
(this=0x7f9a858e4d00) at ../../src/kudu/cfile/cfile_reader.cc:724
This implied that the emplace_back calls in the block decoder were actually
causing reallocation. The issue is that the reserve() was not appropriately
large.
Change-Id: Id3688e2c9e71bd05b5386e939c9582b707e57ff3
---
M src/kudu/cfile/binary_plain_block.cc
1 file changed, 3 insertions(+), 1 deletion(-)
git pull ssh://gerrit.cloudera.org:29418/kudu refs/changes/34/5734/1
--
To view, visit http://gerrit.cloudera.org:8080/5734
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings
Gerrit-MessageType: newchange
Gerrit-Change-Id: Id3688e2c9e71bd05b5386e939c9582b707e57ff3
Gerrit-PatchSet: 1
Gerrit-Project: kudu
Gerrit-Branch: master
Gerrit-Owner: Todd Lipcon <[email protected]>
Gerrit-Reviewer: Adar Dembo <[email protected]>
