feat: DBI column_info, $^S fix, __LINE__ interpolation fix#418
Merged
feat: DBI column_info, $^S fix, __LINE__ interpolation fix#418
Conversation
…re_cached - Implement column_info() using SQLite PRAGMA table_info() to preserve original type case (JDBC uppercases type names, breaking DBIC tests) - Bless metadata result sets (table_info, column_info, etc.) into DBI class with proper Type/Executed/FetchHashKeyName attributes - Intercept literal BEGIN/COMMIT/ROLLBACK SQL in execute() and use JDBC API (setAutoCommit/commit/rollback) instead of executing SQL directly, fixing SQLite JDBC autocommit conflicts - Track AutoCommit state changes when literal transaction SQL is detected - Fix prepare_cached to use per-dbh CachedKids hash instead of global cache, preventing cross-connection cache pollution with :memory: DBs - Handle metadata sth execute() gracefully (no PreparedStatement) - Support PRAGMA pre-fetched rows in fetchrow_hashref DBIx::Class test results: 60/68 active tests pass (88%, was 78%) - t/64db.t: 4/4 real tests pass (was 2/4) - t/752sqlite.t: 34/34 real tests pass (was ~11/34) Generated with [Devin](https://cli.devin.ai/docs) Co-Authored-By: Devin <158243242+devin-ai-integration[bot]@users.noreply.github.com>
…s in eval
When require fails inside eval {}, the __DIE__ handler now sees $^S=1
(inside eval) instead of incorrectly seeing $^S=0 (top-level). The bug
was that evalDepth was decremented by the eval catch block before
catchEval() invoked the handler. Fixed by temporarily incrementing
evalDepth around the handler call in catchEval().
This unblocks t/00describe_environment.t in DBIx::Class which uses
$^S to distinguish eval-guarded optional requires from real crashes.
Generated with [Devin](https://cli.devin.ai/docs)
Co-Authored-By: Devin <158243242+devin-ai-integration[bot]@users.noreply.github.com>
…number
When __LINE__ appeared inside @{[expr]} in double-quoted strings or qr//,
it returned the line number from the start of the enclosing statement
instead of the actual line. This was caused by the inner sub-parser
sharing the outer errorUtil which uses a forward-only cache.
Fixed by:
- Adding baseLineNumber field to Parser for string sub-parsers
- Setting it from the outer source position in StringDoubleQuoted
- Using it in CoreOperatorResolver to compute __LINE__ by counting
inner newlines from the base
This fixes t/106dbic_carp.t tests 2-3 in DBIx::Class which depend on
__LINE__ inside qr// matching the actual source line.
Generated with [Devin](https://cli.devin.ai/docs)
Co-Authored-By: Devin <158243242+devin-ai-integration[bot]@users.noreply.github.com>
- execute_for_fetch now tracks error count and stores [$sth->err, $sth->errstr, $sth->state] in tuple_status on failure (matching DBI 1.647 behavior) - After the loop, dies with error count if RaiseError is on - execute() now sets err/errstr/state on both sth and dbh (was dbh only) - Clears sth error state before each execute call Fixes t/100populate.t test 2 (execute_for_fetch exception detection). Generated with [Devin](https://cli.devin.ai/docs) Co-Authored-By: Devin <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Updated test results after fixing $^S, __LINE__ interpolation, and execute_for_fetch error propagation. Added new "Must Fix" entries for JDBC error message format, SQL formatting differences, and bind parameter attribute handling. Generated with [Devin](https://cli.devin.ai/docs) Co-Authored-By: Devin <158243242+devin-ai-integration[bot]@users.noreply.github.com>
In Perl 5, lexical 'no warnings "redefine"' takes precedence over
the global -w flag ($^W). PerlOnJava was checking $^W with a simple
OR, causing subroutine redefinition warnings even inside no-warnings
blocks. Now checks isWarningDisabled("redefine") first, matching
Perl 5 behavior.
Generated with [Devin](https://cli.devin.ai/docs)
Co-Authored-By: Devin <158243242+devin-ai-integration[bot]@users.noreply.github.com>
…lation When ASM's Frame.merge() crashes on methods with extremely high fan-in (e.g. Sub::Quote-generated subs with 600+ jumps to a single label), EmitterMethodCreator throws InterpreterFallbackException. However, the top-level compileToExecutable() in PerlLanguageProvider didn't catch this exception type, causing the compilation to fail entirely instead of falling back to the interpreter. Add explicit catch for InterpreterFallbackException in compileToExecutable() that returns the pre-built interpreted code from the exception. This fixes DBIx::Class tests that use Sub::Quote (e.g. t/88result_set_column.t). Generated with [Devin](https://cli.devin.ai/docs) Co-Authored-By: Devin <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Two fixes for DBIx::Class compatibility:
1. Implement MODIFY_CODE_ATTRIBUTES call for subroutine attributes.
When a sub is declared with attributes (sub foo : Attr { }),
Perl calls MODIFY_CODE_ATTRIBUTES($package, \&code, @attrs) at
compile time. This was not implemented, causing DBIx::Class
ResultSetManager and any module using code attributes to fail.
Fixes t/40resultsetmanager.t (5/5 planned tests now pass).
2. Fix ROLLBACK TO SAVEPOINT being intercepted as full ROLLBACK.
DBI.java transaction control interceptor matched any SQL starting
with ROLLBACK, including ROLLBACK TO SAVEPOINT. This caused
savepoint operations to incorrectly reset autocommit.
Fixes t/752sqlite.t (171/172 planned tests now pass).
Generated with [Devin](https://cli.devin.ai/docs)
Co-Authored-By: Devin <158243242+devin-ai-integration[bot]@users.noreply.github.com>
When an @inc hook returns a CODE reference (line-reader sub), Perl repeatedly calls it - the sub sets $_ to each line and returns true for more data, or false to stop. This pattern is used by PAR and similar module loaders. Previously only filehandle, scalar ref, and array ref returns from @inc hooks were handled. Now CODE ref returns are properly processed by calling the sub in a loop and accumulating lines from $_. Fixes t/90ensure_class_loaded.t tests 14 and 17 (27/28 now pass). Generated with [Devin](https://cli.devin.ai/docs) Co-Authored-By: Devin <158243242+devin-ai-integration[bot]@users.noreply.github.com>
- 68/314 test programs fully passing, 93.7% individual test pass rate - Documented fixes: -w/redefine, InterpreterFallbackException, MODIFY_CODE_ATTRIBUTES, ROLLBACK TO SAVEPOINT, @inc CODE hooks - Updated remaining failures and next steps list Generated with [Devin](https://cli.devin.ai/docs) Co-Authored-By: Devin <158243242+devin-ai-integration[bot]@users.noreply.github.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Continues DBIx::Class Phase 5 work. Three fixes that improve the test pass rate from 78% to 88%+.
DBI improvements (DBI.java, DBI.pm)
column_info()via SQLitePRAGMA table_info()— preserves original type case (JDBC uppercases type names, breaking DBIC type detection)table_info,column_info,primary_key_info,foreign_key_info,type_inforesults now blessed intoDBIwith proper attributesBEGIN/COMMIT/ROLLBACKrouted through JDBC API (setAutoCommit/commit/rollback) instead of executing SQL directly, fixing SQLite JDBC autocommit conflictsAutoCommitstate tracking —$dbh->{AutoCommit}updated when transaction SQL is detectedprepare_cachedper-dbh cache — changed from global%CACHED_STATEMENTSto per-dbh$dbh->{CachedKids}, preventing cross-connection cache pollution with:memory:databases$^Sfix (WarnDie.java)$^Snow correctly reports 1 inside$SIG{__DIE__}whenrequirefails ineval {}evalDepthwas decremented beforecatchEval()called the handlert/00describe_environment.twhich uses$^Sto distinguish eval-guarded optional requires from real crashes__LINE__interpolation fix (CoreOperatorResolver.java, StringDoubleQuoted.java, Parser.java)__LINE__inside@{[expr]}in double-quoted strings andqr//now returns the correct source lineerrorUtilcachet/106dbic_carp.ttests 2-3 which depend on__LINE__insideqr//matching actual source linesDBIx::Class test results
t/64db.t: 2/4 → 4/4 real testst/752sqlite.t: ~11/34 → 34/34 real testst/00describe_environment.t: crash → fully passingt/106dbic_carp.t: 1/3 → 3/3 real testsTest plan
makepasses (all unit tests)$^Sverified correct for: die in eval, require fail in eval, top-level die__LINE__in@{[]}matches Perl 5 in strings andqr//Generated with Devin