v0.0.215

[Util] Add dereference operator and member access through pointer for…

v0.0.214: [Wasm] Fix `compile_cnf()` for empty CNF.

Solves issue #238 for the `WasmV8` backend. However,
the issue remains open as the interpreter backend
still crashes.

v0.0.213: [CMake] Overhaul V8.cmake

- Use separate output directories based on `CMAKE_BUILD_TYPE`.  This
should fix unexpected rebuilds when working with multiple different
build types.
- Correctly use `CMAKE_SOURCE_DIR` instead of `CMAKE_CURRENT_SOURCE_DIR`
to correctly specify paths.
- Change directory and invoke V8's own `gn` to configure the build.
This seems to be the recommended way of building V8 anyways.  This has
the downside of V8's build files residing in
`third-party/v8/v8/out-<BUILD-TYPE>` rather than in
`build/<BUILD-TYPE>/third-party/V8-src`.  Perhaps the caching in GitLab
CI and CircleCI must be adapted...

v0.0.212: [Doc] Update V8-related documentation.

Mark the manual setup of depot_tools and gn as deprecated.

v0.0.211

Fix #231: Add CI job to test benchmarking

v0.0.210: [Wasm] Fix computation of initial HT capacity.

Up to now, we potentially stored a value larger than a `uint32_t` in
`initial_capacity` leading to undefined behavior. We now properly cast
the value to a `uint64_t` and check if it fits into a `uint32_t`,
otherwise we return the maximal value of `uint32_t`.

Furthermore, we put the repetitive computation of the initial hash table
capacity into its own function.

v0.0.209

[Wasm] Refactor to use `std::nullopt`.

v0.0.208

Fix #230: Fix HyPer connector in benchmarks

v0.0.207: [CI] Sanity check parallel unit test utility

Add a sanity check of the parallel unit test utility to CI.  This check
ensures that the utility correctly reprorts aborts, crashes, failures,
and successes.

Note that the result of this sanity run is added to the artifacts, but
not to the reports.  We do not want to clutter our GitLab test reports.

v0.0.206: [IR] Adapt Cost of Initial Upper Bound to Cost used during Search

During the Search we do not consider the cost for the final result. Currentlty this cost is condidered in the initial upper bound for cost-based pruning.
To resolve this issue, we only consider the sum of the costs of the left and the right subplan as the initial upper bound.
Adapt Test Cases to lower Initial Upper Bound for Pruning.