fix: Hash::MultiValue + blib/arch — splice spill, deref slice, dclone hooks

[fglock]

Summary

Fixes all ./jcpan -t Hash::MultiValue test failures (was 7/10 programs passing, now 10/10), the ./jcpan -t HTTP::Thin blib/arch issue, regex error-handling bugs under JPERL_UNIMPLEMENTED=warn, regex preprocessing issues, and a missing compile-time check for global-only variables in my/state declarations.

Bug 1: JVM backend — splice with constant sub causes ASM frame crash

• handleSpliceBuiltin left the first arg on the JVM operand stack while evaluating remaining args
• When those contained a function call (e.g. constant sub C from use constant), the blockDispatcher GOTOs created inconsistent stack depths at merge points, crashing ASM Frame.merge
• Fix: Register spilling in EmitOperator.java, matching the existing handlePushOperator pattern

Bug 2: Interpreter backend — @$ref[@idx] = ... unsupported

• The array slice assignment handler in CompileAssignment.java only supported @array[@idx] with plain IdentifierNode
• Fix: Added a branch to handle dereferenced array refs using DEREF_ARRAY/DEREF_ARRAY_NONSTRICT opcodes

Bug 3: Storable::dclone — shared refs from STORABLE_freeze hooks

• dclone passed extra refs from STORABLE_freeze directly to STORABLE_thaw without cloning them
• Inside-out objects like Hash::MultiValue ended up sharing internal arrays between original and clone
• Fix: Deep-clone extra refs (indices 1+) before passing to STORABLE_thaw in Storable.java

Bug 4: blib/arch missing in generated Makefile

• ExtUtils::MakeMaker generated a Makefile whose pure_all target never created blib/arch/
• blib.pm requires it and dies if missing
• Fix: Added @mkdir -p blib/arch to the pure_all target in ExtUtils/MakeMaker.pm

Bug 5: JPERL_UNIMPLEMENTED=warn downgrading real regex errors

• ALL regex compilation exceptions were downgraded to warnings, not just unimplemented features
• This caused real errors (e.g. \N{}) to become warnings, allowing execution to continue into infinite loops
• Fix: Only downgrade PerlJavaUnimplementedException, not all exceptions, in RuntimeRegex.java
• Also restructured catch to properly distinguish PerlJavaUnimplementedException (extends PerlCompilerException) from real syntax errors, and wrap Java PatternSyntaxException as downgradable

Bug 6: my $0 / my @_ / my %_ not rejected at compile time

• Perl requires that forced-global variables cannot be lexicalized with my or state
• jperl silently allowed this, which could cause infinite loops
• Fix: Added isGlobalOnlyVariable() check in OperatorParser.addVariableToScope() that emits Can't use global $X in "my" (matching Perl's error message)
• Covers: underscore vars, digit-only names, single ASCII punctuation names, caret/control vars
• Unicode single-char variables (>= 128) are correctly allowed

Bug 7: Regex preprocessor — unimplemented features using wrong exception type

• regexError() (throws PerlCompilerException) was used for unimplemented features like control verbs, (?@...), and lookbehind >255
• These should use regexUnimplemented() (throws PerlJavaUnimplementedException) so they can be downgraded with JPERL_UNIMPLEMENTED=warn
• Fix: Changed these calls to use regexUnimplemented() in RegexPreprocessor.java

Bug 8: handleCodeBlock consuming closing paren meant for handleParentheses

• handleCodeBlock consumed both } and ) of (?{...}), but handleParentheses expects to consume the closing ) itself
• This caused "Unmatched (" errors for (?{...}) code blocks in single-quoted regexes
• Fix: handleCodeBlock now returns offset pointing TO ), letting handleParentheses consume it

Bug 9: handleQuantifier consuming regex metacharacters in brace expressions

• handleQuantifier() used indexOf('}') to find the closing brace, crossing character class and group boundaries
• Pattern like { (?> [^{}]+ | (??{...}) )* } had the (?>... consumed as literal text
• Fix: When braces don't form a valid quantifier, only escape the opening { and return immediately

Bug 10: \x{...} hex escape with non-hex characters crashes

• Integer.parseInt(hexStr, 16) threw fatal NumberFormatException for \x{9bq}
• Fix: Extract valid hex prefix (Perl behavior: stop at first non-hex char). Also handle bare \xNN with non-hex chars by parsing up to 2 hex digits.

Bug 11: NullPointerException when regex fails with JPERL_UNIMPLEMENTED=warn

• regex.patternString was never set in the catch block, causing NPE downstream
• Fix: Set regex.patternString in catch block, add null guard in preProcessRegex()

Bug 12: \p{isAlpha} POSIX-style Unicode property aliases not recognized

• \p{isAlpha}, \p{isSpace} etc. crashed because the Is prefix stripping was case-sensitive
• Fix: Case-insensitive Is/is prefix handling, add Space, Alnum, Punct, White_Space aliases
• This was the return inside a literal can break JVM stack #1 crash blocker for pat.t — unblocked 666 previously unreachable tests

Bug 13: \p{ASCII_Hex_Digit=True} Property=Value syntax not supported

• Property=Value syntax like \p{ASCII_Hex_Digit=True} was passed as a single string to ICU4J
• Fix: Split on = and pass property/value separately. Handle True/False/Yes/No values.

Bug 14: Named capture groups with underscores crash Java regex

• Java regex only allows [a-zA-Z][a-zA-Z0-9]* for group names, but Perl allows \w+
• (?<_>abc) and (?<foo_bar>abc) caused fatal PatternSyntaxException
• Fix: Encode underscores as "U95" in Java regex names (_ → U95, foo_bar → fooU95bar), decode back when accessing %+/%- hashes, handle \k<name> backrefs

Test plan

• make clean ; make passes (all unit tests)
• ./jcpan -t Hash::MultiValue — 10/10 test programs, 55/55 subtests PASS
• ./jcpan -t HTTP::Thin — blib/arch created correctly
• op/local.t — no regression vs master (142/170 ok on both)
• my $0, my @_, my %_, my $!, my $/, my $^W, state $_ all correctly rejected
• our $_, local $_, my $a, my %ENV all correctly allowed
• uni/variables.t — 66880/66880 (matches master)
• re/pat_rt_report.t — 2431/2515 (improved from 2397)
• re/pat.t — 1076/1298, all 1298 now run (was 533/632, crashed mid-run)
• re/pat_advanced.t — 731/838 (improved from 63)
• re/reg_eval_scope.t — 7/49 (improved from 0)

Generated with Devin