Previously we would calculate the index of the first parent node as
heap.size() (which is initialized to non_zero_freqs), so in the edge
case in which all symbols had a non-zero frequency, we would use the
Size-index entry in the array for both the first symbol's leaf node,
and the first parent node.
The result would either be a non-optimal huffman code (bad), or an
illegal huffman code that would then go on to crash due to an error
check in CanonicalCode::from_bytes. (worse)
We now store parent nodes starting at heap.size() - 1, which eliminates
the potential overlap, and resolves the issue.
I'm not entirely sure why this is needed, but it's the same ol'
workaround we're using in a bazillion places where we get caught trying
to do JavaScripty things without a running execution context.
This ensures that repeated loads of the same module succeed. (There is a
specific criteria where the same exact module object has to be returned
for multiple loads of the same referrer + specifier.)
Note that we don't check the referrer at the moment, that's a FIXME.
In particular, this patch focuses on:
- Updating the old "import assertions" to the new "import attributes"
- Allowing realms as module import referrer
This allows them to participate in the ownership graph and fixes a
lifetime issue in module loading found by ASAN.
Co-Authored-By: networkException <networkexception@serenityos.org>
This fixes an issue where we end up in a state where we have no
execution context + a main thread event loop with an empty incumbent
settings object stack.
We were previously only returning the controllers current
[[byobRequest]] instead of taking into account pending pull intos.
Rename the getter function which would return the controllers
[[byobRequest]] slot to `raw_byob_request` to differentiate it from
the IDL getter.
This also leaves a FIXME for a spec step which we are also not currently
implementing correctly.
By default the bindings go to /usr/local on the host, which is a very
big no-no; this path is not affected by CMAKE_INSTALL_PREFIX, so this
commit sets the LLVM_OCAML_INSTALL_PATH variable instead.
It should be noted that disabling the ocaml bindings doesn't make all
the users of this variable go away, so this commit doesn't do so.
This also sorts the -DFOO options passed to cmake, because...sorting.
We forgot to reset all the variables that keep track of suggestion
state, resulting in an underflow value when calculating the lines to
display completion suggestions later.
Setting `m_times_tab_pressed` to 0 apparently forces it to recalculate
the those variables and seems to fix the problem.
Fixes#22128
