PR #48154 introduced a sanity check in the form of a debug assertion
that the input values for VarInt encoding are not too big. Such values
should be exceptionally rare in practice but the AST fuzzer managed to
trigger the assertion regardless. The strategy to deal with such values
until now was to bypass the check by limiting the value to the maximum
allowed value (see #48412). Because a new AST Fuzzer failure appeared
(#48497) and there may be more failures in future, this PR changes the
sanity check from an assert to an exception.
Fixes: #48497
- Introduced with the C++20 <bit> header
- The problem with __builtin_c(l|t)z() is that 0 as input has an
undefined result (*) and the code did not always check. The std::
versions do not have this issue.
- In some cases, we continue to use buildin_c(l|t)z(), (e.g. in
src/Common/BitHelpers.h) because the std:: versions only accept
unsigned inputs (and they also check that) and the casting would be
ugly.
(*) https://gcc.gnu.org/onlinedocs/gcc/Other-Builtins.html
There are lots of places where HashTable is used with
AllocatorWithStackMemory, but the size of allocator inline memory is set
incorrectly, and it's less than the initial HashTable buffer size.
Because of this, HashTable is always allocated on the heap, and the
inline memory becomes a useless dead weight.
For PODArray, we have previously added a helper template that makes sure
these values are in sync, so there was only one such discrepancy left,
in the unit test.
