Use C99 designated initialisers for PerlIO_funcs

[ilmari]

This avoids having to count NULLs for layers that only implement a subset of functions.

---

• This set of changes does not require a perldelta entry.

[tonycoz]

I think it is a silly warning with designated initializers, but g++ produces a mess of warnings:

g++ -c -DPERL_CORE -fwrapv -DDEBUGGING -fno-strict-aliasing -pipe -fstack-protector-strong -I/usr/local/include -D_LARGEFILE_SOURCE -D_FILE_OFFSET_BITS=64 -O2 -g -Wall -Werror=pointer-arith -Werror=vla -Wextra -Wno-long-long -Wno-nonnull-compare -Wwrite-strings -Wno-use-after-free perlio.c
perlio.c:1031:1: warning: missing initializer for member ‘_PerlIO_funcs::size’ [-Wmissing-field-initializers]
 1031 | };
      | ^
perlio.c:1031:1: warning: missing initializer for member ‘_PerlIO_funcs::Popped’ [-Wmissing-field-initializers]
perlio.c:1031:1: warning: missing initializer for member ‘_PerlIO_funcs::Binmode’ [-Wmissing-field-initializers]
perlio.c:1031:1: warning: missing initializer for member ‘_PerlIO_funcs::Getarg’ [-Wmissing-field-initializers]
... and many more

Adding C++ default initializers prevents the warnings but that would be a mess of macros for us.

I suspect the best choice would be to suppress the warning:

GCC_DIAG_IGNORE_DECL(-Wmissing-field-initializers);
PERLIO_FUNCS_DECL(PerlIO_remove) = {
     .fsize = sizeof(PerlIO_funcs),
    .name = "pop",
    .kind = PERLIO_K_DUMMY | PERLIO_K_UTF8,
    .Pushed = PerlIOPop_pushed,
    .Open = PerlIOBase_open,
};
GCC_DIAG_RESTORE

[leonerd]

I encountered this problem of the g++ warnings again recently in #24508, and discussed it out loud on the mailing list https://www.nntp.perl.org/group/perl.perl5.porters/2026/06/msg271009.html

While it is technically possible to silence the warnings by wrapping them as suggested, I would find this annoyingly verbose to have to remember to do for every C99-style struct initialisation we ever write into our code, just to quiet a warning in the tiny corner-case of us using g++ to compile a C program. I'm not going to veto it for this specific PR, but I am keen to look into finding a better solution to the overall problem, as I mentioned on the mailing list.

[ilmari]

Would it work to wrap the suppression in a macro?

#define PERLIO_FUNCS_DEF(funcs, ...) \
    GCC_DIAG_IGNORE_DECL(-Wmissing-field-initializers); \
    PERLIO_FUNCS_DECL(funcs) = { __VA_ARGS__ };         \
    GCC_DIAG_RESTORE

and then do:

PERLIO_FUNCS_DEF(PerlIO_remove,
     .fsize = sizeof(PerlIO_funcs),
    .name = "pop",
    .kind = PERLIO_K_DUMMY | PERLIO_K_UTF8,
    .Pushed = PerlIOPop_pushed,
    .Open = PerlIOBase_open,
);

Ore even a generic STRUCT_INIT macro?

#define STRUC_INIT(decl, ...) \
    GCC_DIAG_IGNORE_DECL(-Wmissing-field-initializers); \
    decl = { __VA_ARGS__ };         \
    GCC_DIAG_RESTORE

[leonerd]

As the g++ warnings are a wider issue that affects a bunch of PRs and potential future code, and also is already an issue in blead right now, I've opened a separate issue #24553 to discuss it

[leonerd]

Now that #24559 is in, this shouldn't cause upsets to g++