…acters

…acters

- Decimal.cs: the new consumed-count TryParse overloads now use ValidateParseStyleDecimal instead of ValidateParseStyleFloatingPoint, matching the rest of decimal parsing (which rejects hex/binary specifiers).

- INumberBase.cs: the default consumed-count TryParse implementations no longer strip NumberStyles.AllowTrailingInvalidCharacters before delegating. Per the approved design, the style is passed through unmodified so a type that does not understand the flag throws as expected, rather than silently succeeding with a masked style.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

…ests

- Allow NumberStyles.AllowTrailingInvalidCharacters to be combined with
  HexFloat. ValidateParseStyleFloatingPoint previously rejected
  HexFloat | AllowTrailingInvalidCharacters (via the `style & ~HexFloat`
  mask), which made the AllowTrailingInvalidCharacters handling already
  present in TryParseHexFloatingPoint unreachable. Widen the mask to
  permit the flag, matching how ValidateParseStyleInteger already permits
  HexNumber | AllowTrailingInvalidCharacters and
  BinaryNumber | AllowTrailingInvalidCharacters.

- Update the AllowBinarySpecifier XML docs to note that
  AllowTrailingInvalidCharacters may also be combined with it, consistent
  with the updated validation behavior.

- Add UTF-8 (ReadOnlySpan<byte>) failure-path assertions
  (bytesConsumed == 0) to Parse_AllowTrailingInvalidCharacters_Invalid in
  the test files that were missing them (Double, Single already had it;
  Half, Decimal, Decimal32/64/128, and Int64 did not), locking down the
  new consumed-count contract for the UTF-8 overloads.

- Add double hex-float regression cases for
  HexFloat | AllowTrailingInvalidCharacters, including one that stops on a
  trailing invalid character after consuming trailing whitespace.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

…F-8 invalid assertion

- Use CultureInfo.InvariantCulture instead of null provider for floating-point/decimal
  Parse_AllowTrailingInvalidCharacters test data (Double/Single/Half/Decimal) so cases
  with a decimal point/exponent are deterministic across cultures.
- Assert the UTF-8 (bytesConsumed == 0) failure contract in BigInteger's
  Parse_AllowTrailingInvalidCharacters_Invalid test.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

…, expand special-value tests

- Remove [MethodImpl(NoInlining)] from ConsumeTrailingNulls; it is small,
  centralized, and always called on the trailing path.
- Restructure TryParseFloat so the special-value (Infinity/NaN) handling runs
  for both the hex and non-hex paths. Hex floating-point now supports the
  Infinity/NaN symbols and the misleading "not supported" comment is corrected.
- Add the surrounding-whitespace special-value (Infinity/NaN) test cases to
  Single/Half/Decimal32/Decimal64 to match Double/Decimal128.
- Add hex-float Infinity/NaN test cases (valid and trailing-invalid) to
  Double/Single/Half and remove the now-obsolete invalid entries that asserted
  NaN/Infinity were rejected under HexFloat.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

Matching the Infinity/NaN symbols via SpanStartsWith(..., OrdinalIgnoreCase)
routed the UTF-8 path through StartsWithIgnoreCaseUtf8_Scalar, which trips a
Debug.Assert(length == 0) when the trailing 1-3 byte chunk matches case-
insensitively but not exactly (e.g. UTF-8 "NAN" vs the "NaN" symbol). The
UTF-16 tests only used exact-case symbols, so this surfaced via case-variant
UTF-8 input such as the System.Text.Json named-literal parsing.

Match the symbol against an equal-length slice using EqualsOrdinalIgnoreCase
(the well-tested equality path, as the original code did) which is
semantically identical to a case-insensitive prefix match here.

Also remove the now-obsolete "NaN/Infinity not valid for HexFloat" invalid
entries from BFloat16Tests and add the corresponding valid hex Infinity/NaN
cases, matching Double/Single/Half. Add mixed-case special-value cases to
DoubleTests to exercise the UTF-8 case-insensitive path.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>