Dan Luu
Dan Luu
In ConditionalNode.h ``` // TODO: Evaluating both expressions in advance of the test is // sub-optimal, but it is currently required to guarantee consistent // state: the execution in NativeJIT...
Register.h ``` // TODO: Need to avoid "static initialization order fiasco" for register definitions. // See http://www.parashift.com/c++-faq/static-init-order.html. // Plan is to use constexpr when VS2013 becomes available to Bing build....
It would be nice if ReturnNode::CompileAsRoot could generate into RAX instead of moving. Consider example of a tree that is just Return(Immediate(5)). Right now this generates ``` mov r15, 5...
ExceptionBuffer.cpp throws an exception in its destructor.
``` ==23425== 416 bytes in 4 blocks are definitely lost in loss record 2 of 3 ==23425== at 0x4C2E0EF: operator new(unsigned long) (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so) ==23425== by 0x599E98: BitFunnel::Shard::CreateNewActiveSlice() (Shard.cpp:134) ==23425==...
~~~ Processing query " blood" ~~~ This makes it look like the query is one the term " blood", not the term "blood". This seems to happen to all terms.
~~~ 0: failOnException 1: cache chunk sonnet0 2: 3: Processing query " blood" True positive count: 2 False positive count: 2 False negative count: 0 True positives: 2, 11 False...
A false negative causes verify to kill the test with an exception. Because the tests do `show rows` after `verify`, you don't get the rows output conveniently printed.
This has ~~~ std::unique_ptr stream( new std::ofstream(filename, mode)); stream->exceptions(std::ofstream::badbit | std::ofstream::badbit); ~~~ I wonder if one of these `std::ofstream::badbit`s was intended to be `std::ofstream::failbit`? Later in the same file, `OpenForRead`...
We've had some tests fail because they want to allocate more memory than Travis give us. I've fixed this for now by cutting down the size of buffers that we...