[IO::Handle] Destroying a borrowed DATA handle closes the stream
Mantenedores costumam responder em até 1 dia
Ninguém assumiu esta issue ainda.
Avaliação
- Dificuldade
- 4/5
- Tempo estimado
- 3-5 dias
- Facilidade para iniciantes
- 45/100
Direção de pesquisa
Start by reading IO::Handle::DESTROY and the overlapping RuntimeIO ownership work in #1649 and PR #1938; coordinate before changing handle cleanup. Add the focused regression for destroying a wrapper around a borrowed package DATA handle, first validating it with system Perl. Done means the borrowed handle remains readable, owned descriptors and explicit close still clean up, and the regression plus CPAN::Repository 0.010 suite pass on both backends.
Escrita pelo modelo de indexação a partir do texto da issue.
Descrição
Summary
PerlOnJava closes a package's borrowed DATA handle when a temporary FileHandle wrapper is destroyed. File::MMagic uses this pattern for its built-in magic table. After the first archive type check, the package DATA stream is closed; subsequent checks load no magic entries and return application/octet-stream. Archive::Any has no handler for that type, so Dist::Data then calls extract on an undefined archive object.
This caused CPAN::Repository 0.010 to fail in CPAN random run 20261111-090841-71684, test t/10-simple.t.
Reproducer
With File::MMagic 1.30 and two gzip archives, run both checks in the same process:
use File::MMagic;
for my $path (@ARGV) {
print File::MMagic->new()->checktype_filename($path), "\n";
}
Use the CPAN::Repository 0.010 fixtures My-Sample-Distribution-0.003.tar.gz and My-Other-Sample-0.001.tar.gz as the two arguments.
Expected on system Perl:
application/x-gzip
application/x-gzip
Actual on PerlOnJava JVM and interpreter:
application/x-gzip
application/octet-stream
The archived t/10-simple.t failure then reports:
No handler available for type 'application/octet-stream'
Can't call method "extract" on an undefined value
Root cause evidence
File::MMagic::new borrows *File::MMagic::DATA{IO} and blesses that handle as a FileHandle. PerlOnJava's IO::Handle::DESTROY unconditionally closes a handle with a defined fileno, without distinguishing a borrowed package handle from a handle owned by that wrapper. Instrumentation shows that after the temporary detector is released, the DATA handle is closed (tell returns -1 and eof is true); the next detector loads no magic entries. Keeping a detector object alive across both calls avoids the failure.
Validation and environment
- CPAN::Repository 0.010: PerlOnJava JVM and interpreter both fail
t/10-simple.twith this error. - System Perl 5.34.1: all four upstream test files pass, 16 tests total.
File::MMagic1.30 andMIME::Types2.30 were installed in an isolated temporary library because the host Perl lacked them. - PerlOnJava source:
308a379fc6778e5db851d10c260656e82ce03944; OpenJDK 24.0.2, macOS 26.7.1 arm64. - The target distribution and the relevant archive dependencies are pure Perl; this is not an XS build or environment-prerequisite failure.
Acceptance criteria
- Add a focused project-owned regression showing that destroying a wrapper for a borrowed package
DATAhandle leaves that handle readable, validated first with system Perl. - Preserve cleanup for genuinely owned file descriptors and explicit
closebehavior. - Pass the focused regression on JVM and interpreter, and pass the CPAN::Repository 0.010 upstream suite on both backends.
Coordination and related issues
The active #1649 ownership work and PR #1938 cover broad IO resource ownership and modify RuntimeIO; coordinate before editing overlapping handle cleanup paths. #1244 records a different compile-time-loading failure that also affected a DATA handle. #2024/#2035 addressed ambiguous close warnings, not this ownership bug.
- Linguagem predominante
- Perl
- Estrelas
- 67
- Forks
- 8
- Merge médio
- 5h 12min
- PRs com merge (30d)
- 131
Preparar o ambiente
- Inclui um Dockerfile ou arquivo Docker Compose
- Sem modelo de pull request
- Ler o guia de contribuição
Primeiros passos
- Leia a issue inteira e depois o guia de contribuição do projeto.
- Comente na issue dizendo que vai assumir — evita que duas pessoas façam o mesmo trabalho.
- Faça um fork do repositório e trabalhe em uma branch.
- Abra um pull request que referencie o número da issue.
Mais de fglock/PerlOnJava
-
area:cpan-port bug
Dificuldade 2/5 1-3 horas Facilidade para iniciantes 70/100
fglock/PerlOnJava#2153 ·
Mantenedores costumam responder em até 1 dia
-
area:cpan-port area:parser bug
Dificuldade 2/5 1-3 horas Facilidade para iniciantes 68/100
fglock/PerlOnJava#2137 ·
Mantenedores costumam responder em até 1 dia
-
area:backend area:cpan-port area:runtime bug
Dificuldade 2/5 1-3 horas Facilidade para iniciantes 76/100
fglock/PerlOnJava#2071 ·
Mantenedores costumam responder em até 1 dia
-
area:runtime documentation enhancement
Dificuldade 2/5 1-3 horas Facilidade para iniciantes 72/100
fglock/PerlOnJava#1579 ·
Mantenedores costumam responder em até 1 dia
-
area:cpan-port bug
Dificuldade 3/5 1-2 dias Facilidade para iniciantes 62/100
fglock/PerlOnJava#2157 ·
Mantenedores costumam responder em até 1 dia
Todas as issues de fglock/PerlOnJava
Issues semelhantes
-
Dev-output directory preparation accepts file-valued lib/include paths and fails on dangling linksAbertateam-cli
Dificuldade 2/5 1-3 horas Facilidade para iniciantes 70/100
Mantenedores costumam responder em até 2 dias
-
Dificuldade 2/5 1-3 horas Facilidade para iniciantes 66/100
-
Dificuldade 1/5 Menos de uma hora Facilidade para iniciantes 93/100
Mantenedores costumam responder em até 1 dia
-
Dificuldade 2/5 1-3 horas Facilidade para iniciantes 73/100
inverse-inc/packetfence#9728 ·
Mantenedores costumam responder em até 1 dia
-
fix(char-create): show the real reason a character is refusedTalvez já em andamento @baphomello assumiu há 1 dia. Aberta
Dificuldade 2/5 1-3 horas Facilidade para iniciantes 84/100
Mantenedores costumam responder em até 1 dia