Which libc/stdio IO functions Perl uses to open a file?

Question

Which libc/stdio functions does Perl use to open and read a file (e.g. for perl -e 'open(F,"<","test.txt");print(<F>);')?

I tried running this command under strace and it gave me a sequence open -> fcntl -> ioctl -> lseek -> fstat -> mmap -> read. strace intercepts syscalls, so I'm wondering what are the higher-level libc/stdio functions are actually being called in https://github.com/Perl/perl5/blob/blead/perlio.c and https://github.com/Perl/perl5/blob/blead/doio.c? The Perl I/O codebase has many layers of indirection, so I'm struggling to understand it.

I'm not a Perl pro, so not understanding https://perldoc.perl.org/PerlIO and the codebase sufficiently enough :(

Thanks!

Answer 1

Depends on the layer.

There are two options for the base layer of ordinary handles: :unix and :stdio.

• The :unix layer is is defined in perlio.c as PerlIO_unix.

  • For opening, :unix uses PerlIOUnix_open, which calls PerlLIO_open3_cloexec, which calls PerlLIO_open3, which is defined as open (except maybe on MVS).

  • For writing, :unix uses PerlIOUnix_write, which calls PerlLIO_write, which is defined as write.

• The :stdio layer is is defined in perlio.c as PerlIO_stdio.

  • For opening, :stdio uses PerlIOStdio_open. Follow from there.

  • For writing, :stdio uses PerlIOStdio_write, which calls PerlSIO_fwrite, which is defined as fwrite.

The selection of the base layer appears to be:

1. If a layer is provided to open, use that one.
2. If a process-wide default is provided (e.g. using the PERLIO env var), use that one.
3. Otherwise, use the default selected when Perl was built. This can apparently be :stdio in some circumstances, but it's usually :perlio, and :perlio adds an underlying :unix layer.

Note that :unix is used on Windows as well. It refers to the libc functions rather than anything specific to Unix.

[All links for v5.40.0.]