C++ builds fail with modern C++ compilers #21073

tonycoz · 2023-05-04T01:55:05Z

Module: Compress::Raw::Bzip2

Other modules may also be a problem.

Description

With a modern C++ compiler supporting and defaulting to C++ 17, the register storage class modifier is no longer supported, which causes the build to fail:

 c++ -DNO_MATHOMS -I/usr/local/include -D_LARGEFILE_SOURCE -D_FILE_OFFSET_BITS=64 -Wall -Werror=pointer-arith -Werror=vla -Wextra -Wno-long-long -Wno-declaration-after-statement -Wc++-compat -Wwrite-strings -O2   -DVERSION=\"2.204\" -DXS_VERSION=\"2.204\" -fPIC "-I../.."  -DBZ_NO_STDIO  compress.c
compress.c:190:13: error: ISO C++17 does not allow 'register' storage class specifier [-Wregister]
            register UChar  rtmp;
            ^~~~~~~~~

Note that at least one C99 construct we use isn't valid in C++:

In file included from globals.c:25:
In file included from ./perl.h:6191:
./opcode.h:1004:2: warning: array designators are a C99 extension [-Wc99-designator]
        [OP_NULL]               = Perl_pp_null,
        ^~~~~~~~~

(designated struct initialization is supported in C++20 but not arrays)

Steps to Reproduce

# requires clang-16, installed here from the llvm deb archive
./Configure -des -Dusedevel -Dcc=clang++-16 -Accflags='-g -fno-omit-frame-pointer -fsanitize=address -fno-common -x c++' -Aldflags=-fsanitize=address -Accflags=-DNO_MATHOMS && make

Expected behavior

Successful build.

Workaround

Build with -std=c++11, which changes the message to a warning:

clang++-16 -c  -I./zlib-src -g -fno-omit-frame-pointer -fsanitize=address -fno-common -x c++ -std=c++11 -DNO_MATHOMS -I/usr/local/include -D_LARGEFILE_SOURCE -D_FILE_OFFSET_BITS=64 -Wall -Werror=pointer-arith -Werror=vla -Wextra -Wno-long-long -Wno-declaration-after-statement -Wc++-compat -Wwrite-strings -O2   -DVERSION=\"2.204\" -DXS_VERSION=\"2.204\" -fPIC "-I../.."  -DNO_VIZ -DZ_SOLO   -DZ_PREFIX  -DGZIP_OS_CODE=3  -DPerl_crz_BUILD_ZLIB=1 deflate.c
deflate.c:1282:5: warning: 'register' storage class specifier is deprecated and incompatible with C++17 [-Wdeprecated-register]
    register Bytef *scan = s->window + s->strstart; /* current string */
    ^~~~~~~~~

Perl configuration

# from miniperl
Summary of my perl5 (revision 5 version 37 subversion 12) configuration:
  Commit id: 6bbe21a508a82ef4a3849bd69a435b3cf8d5bf40
  Platform:
    osname=linux
    osvers=5.10.0-21-amd64
    archname=x86_64-linux
    uname='linux venus 5.10.0-21-amd64 #1 smp debian 5.10.162-1 (2023-01-21) x86_64 gnulinux '
    config_args='-des -Dusedevel -Dcc=clang++-16 -Accflags=-g -fno-omit-frame-pointer -fsanitize=address -fno-common -x c++ -Aldflags=-fsanitize=address -Accflags=-DNO_MATHOMS'
    hint=recommended
    useposix=true
    d_sigaction=define
    useithreads=undef
    usemultiplicity=undef
    use64bitint=define
    use64bitall=define
    uselongdouble=undef
    usemymalloc=n
    default_inc_excludes_dot=define
  Compiler:
    cc='clang++-16'
    ccflags ='-g -fno-omit-frame-pointer -fsanitize=address -fno-common -x c++ -DNO_MATHOMS -I/usr/local/include -D_LARGEFILE_SOURCE -D_FILE_OFFSET_BITS=64'
    optimize='-O2'
    cppflags='-g -fno-omit-frame-pointer -fsanitize=address -fno-common -x c++ -DNO_MATHOMS -I/usr/local/include'
    ccversion=''
    gccversion='Debian Clang 16.0.3 (++20230502062952+da3cd333bea5-1~exp1~20230502183046.82)'
    gccosandvers=''
    intsize=4
    longsize=8
    ptrsize=8
    doublesize=8
    byteorder=12345678
    doublekind=3
    d_longlong=define
    longlongsize=8
    d_longdbl=define
    longdblsize=16
    longdblkind=3
    ivtype='long'
    ivsize=8
    nvtype='double'
    nvsize=8
    Off_t='off_t'
    lseeksize=8
    alignbytes=8
    prototype=define
  Linker and Libraries:
    ld='clang++-16'
    ldflags =' -fsanitize=address -L/usr/local/lib'
    libpth=/usr/lib/llvm-16/lib/clang/16/lib /usr/local/lib /usr/lib/x86_64-linux-gnu /usr/lib /usr/lib64
    libs=-lpthread -lgdbm -ldl -lm -lcrypt -lutil -lc -lgdbm_compat
    perllibs=-lpthread -ldl -lm -lcrypt -lutil -lc
    libc=libc-2.31.so
    so=so
    useshrplib=false
    libperl=libperl.a
    gnulibc_version='2.31'
  Dynamic Linking:
    dlsrc=dl_dlopen.xs
    dlext=so
    d_dlsymun=undef
    ccdlflags='-Wl,-E'
    cccdlflags='-fPIC'
    lddlflags='-shared -O2 -L/usr/local/lib'

Characteristics of this binary (from libperl): 
  Compile-time options:
    HAS_LONG_DOUBLE
    HAS_STRTOLD
    HAS_TIMES
    NO_MATHOMS
    PERLIO_LAYERS
    PERL_COPY_ON_WRITE
    PERL_DONT_CREATE_GVSV
    PERL_EXTERNAL_GLOB
    PERL_HASH_FUNC_SIPHASH13
    PERL_HASH_USE_SBOX32
    PERL_IS_MINIPERL
    PERL_MALLOC_WRAP
    PERL_OP_PARENT
    PERL_PRESERVE_IVUV
    PERL_USE_DEVEL
    PERL_USE_SAFE_PUTENV
    USE_64_BIT_ALL
    USE_64_BIT_INT
    USE_LARGE_FILES
    USE_LOCALE
    USE_LOCALE_COLLATE
    USE_LOCALE_CTYPE
    USE_LOCALE_NUMERIC
    USE_LOCALE_TIME
    USE_PERLIO
    USE_PERL_ATOF
    USE_SITECUSTOMIZE
  Built under linux
  Compiled at May  4 2023 10:48:51
  %ENV:
    PERLBREW_BASHRC_VERSION="0.43"
    PERLBREW_HOME="/home/tony/.perlbrew"
    PERLBREW_MANPATH=""
    PERLBREW_PATH="/home/tony/perl5/perlbrew/bin"
    PERLBREW_ROOT="/home/tony/perl5/perlbrew"
    PERLBREW_VERSION="0.67"
  @INC:
    /home/tony/dev/perl/git/perl6/cpan/AutoLoader/lib
    /home/tony/dev/perl/git/perl6/dist/Carp/lib
    /home/tony/dev/perl/git/perl6/dist/PathTools
    /home/tony/dev/perl/git/perl6/dist/PathTools/lib
    /home/tony/dev/perl/git/perl6/cpan/ExtUtils-Install/lib
    /home/tony/dev/perl/git/perl6/cpan/ExtUtils-MakeMaker/lib
    /home/tony/dev/perl/git/perl6/cpan/ExtUtils-Manifest/lib
    /home/tony/dev/perl/git/perl6/cpan/File-Path/lib
    /home/tony/dev/perl/git/perl6/ext/re
    /home/tony/dev/perl/git/perl6/dist/Term-ReadLine/lib
    /home/tony/dev/perl/git/perl6/dist/Exporter/lib
    /home/tony/dev/perl/git/perl6/ext/File-Find/lib
    /home/tony/dev/perl/git/perl6/cpan/Text-Tabs/lib
    /home/tony/dev/perl/git/perl6/dist/constant/lib
    /home/tony/dev/perl/git/perl6/cpan/version/lib
    /home/tony/dev/perl/git/perl6/cpan/Getopt-Long/lib
    /home/tony/dev/perl/git/perl6/cpan/Text-ParseWords/lib
    /home/tony/dev/perl/git/perl6/cpan/ExtUtils-PL2Bat/lib
    /home/tony/dev/perl/git/perl6/lib

The text was updated successfully, but these errors were encountered:

tonycoz · 2023-05-04T02:21:48Z

While we could fix this particular failure by asking upstreams not to use register, it doesn't fix the more general problem that C isn't C++: the designated array initializer means we can't perform similar testing with C++ compilers don't provide this extension.

I remember two reasons that have been mentioned for testing with C++:

testing that our headers are C++ compatible
stricter type checking from C++

I think 1. is useful, but rather than requiring that the whole of perl build with C++, I think it would be better to limit that to the headers, this reduces the load on our upstreams and lets us use C features in most code without the worry that they aren't supported in C++.

I think the stricter type checking is close to nonsense, the only stricter type check I'm aware of for our purposes is the requirement to cast void * to other types, the other types of strictness only apply when using C++ style casts.

khwilliamson · 2023-05-07T01:13:19Z

I do my development using g++. This is a habit from when C++ was new, and was billed as a better C, even if one didn't use the OO capabilities, and the C++ compiler was merely a preprocessor that translated the source to C (Cfront). Since then the languages have diverged more. I don't know the current state of the advantages of C++ compilers over C, if any.

tonycoz · 2023-05-08T00:17:54Z

and was billed as a better C, even if one didn't use the OO capabilities

This can be true, though I think it mostly depended on using C++ features (like new and delete).

Some of the problem with C vs C++ is some compilers only warn on code the C standard considers ill-formed code, like passing an int * when a STRLEN * is required, while I recall the Turbo C compiler I started with would throw an error on such code (but it's been a long time.)

gcc warns on such code, g++ throws an error, even though it's ill-formed in both languages.

But in terms of progress on this, I can see us making a local change to remove register and submit it upstream, but it's not going to fix the other issue.

The win32 Makefiles' USE_CPLUSPLUS support appears to be broken, but if I manually supply -TP I get:

        cl -c -I.. -nologo -GF -W3 -MD -I..\lib\CORE -I.\include -I. -I.. -DWIN32 -D_CONSOLE -DNO_STRICT -DWIN64 -D_CRT_SECURE_NO_DEPRECATE -D_CRT_NONSTDC_NO_DEPRECATE -D_WINSOCK_DEPRECATED_NO_WARNINGS -DPERLDLL -DPERL_CORE   -O1 -Zi -GL -fp:precise -TP -DPERL_TEXTMODE_SCRIPTS -DMULTIPLICITY -DPERL_IMPLICIT_SYS -Fo..\globals.obj ..\globals.c
globals.c
C:\Users\Tony\dev\perl\git\perl\opcode.h(571): error C3480: 'OP_NULL': a lambda capture must be a variable from an enclosing function scope
C:\Users\Tony\dev\perl\git\perl\opcode.h(571): error C3260: '=': skipping unexpected token(s) before lambda body
C:\Users\Tony\dev\perl\git\perl\opcode.h(996): warning C4467: usage of ATL attributes is deprecated
C:\Users\Tony\dev\perl\git\perl\opcode.h(996): error C2337: 'OP_NULL': attribute not found
C:\Users\Tony\dev\perl\git\perl\opcode.h(573): error C2059: syntax error: '='
C:\Users\Tony\dev\perl\git\perl\opcode.h(996): error C2059: syntax error: ';'
C:\Users\Tony\dev\perl\git\perl\opcode.h(1425): error C3480: 'OP_NULL': a lambda capture must be a variable from an enclosing function scope
C:\Users\Tony\dev\perl\git\perl\opcode.h(1425): error C3260: '=': skipping unexpected token(s) before lambda body
C:\Users\Tony\dev\perl\git\perl\opcode.h(1850): error C2337: 'OP_NULL': attribute not found
C:\Users\Tony\dev\perl\git\perl\opcode.h(1428): error C2059: syntax error: '='
C:\Users\Tony\dev\perl\git\perl\opcode.h(1850): error C2059: syntax error: ';'
C:\Users\Tony\dev\perl\git\perl\opcode.h(2274): error C3480: 'OP_NULL': a lambda capture must be a variable from an enclosing function scope
C:\Users\Tony\dev\perl\git\perl\opcode.h(2274): error C3260: '=': skipping unexpected token(s) before lambda body
..\globals.c(45): fatal error C1004: unexpected end-of-file found
N

gcc vs g++

tony@venus:.../perl/git$ gcc -c pointers.c 
pointers.c: In function ‘g’:
pointers.c:6:7: warning: passing argument 1 of ‘f’ from incompatible pointer type [-Wincompatible-pointer-types]
    6 |     f(&t);
      |       ^~
      |       |
      |       size_t * {aka long unsigned int *}
pointers.c:3:15: note: expected ‘int *’ but argument is of type ‘size_t *’ {aka ‘long unsigned int *’}
    3 | extern void f(int *);
      |               ^~~~~
tony@venus:.../perl/git$ g++ -c pointers.c 
pointers.c: In function ‘void g(size_t)’:
pointers.c:6:7: error: cannot convert ‘size_t*’ {aka ‘long unsigned int*’} to int*’
    6 |     f(&t);
      |       ^~
      |       |
      |       size_t* {aka long unsigned int*}
pointers.c:3:15: note:   initializing argument 1 of ‘void f(int*)’
    3 | extern void f(int *);
      |               ^~~~~

Leont · 2023-05-08T15:05:12Z

The win32 Makefiles' USE_CPLUSPLUS support appears to be broken, but if I manually supply -TP I get:

Then we should possibly just revert ab28d72. As far as I can tell it doesn't really do anything functionally anyway.

khwilliamson · 2023-05-11T14:09:32Z

@Leont's idea seems reasonable to me

Leont · 2023-05-12T11:20:27Z

But in terms of progress on this, I can see us making a local change to remove register and submit it upstream, but it's not going to fix the other issue.

Adding a strategically placed #define register may be the easiest solution until upstream updates their code.

leonerd · 2023-05-12T15:29:21Z

There's two separate problems:

register
Use of C99 designated initialisers

We could just revert ab28d72 to undo the initialisers, but that still leaves register. I don't see a good reason not to revert that one anyway, at least it half-fixes the issue.

leonerd · 2023-05-12T15:34:58Z

Ahh except it seems a trivial revert isn't sufficient because it conflicts. and in any case that original commit looks like it should have changed opcode.h. et.al. but they don't appear modified in the same commit. Some more investigation required - @Leont do you want to look into it? Maybe just undo the [NAME] = ... part of the code, as it seems to have kept the right enum order

thesamesam · 2023-05-13T05:31:12Z

While we could fix this particular failure by asking upstreams not to use register, it doesn't fix the more general problem that C isn't C++: the designated array initializer means we can't perform similar testing with C++ compilers don't provide this extension.
[...]
I think 1. is useful, but rather than requiring that the whole of perl build with C++, I think it would be better to limit that to the headers, this reduces the load on our upstreams and lets us use C features in most code without the worry that they aren't supported in C++.

This is what most projects I'm familiar with do. systemd has a batch of tests for this, for example.

I think the stricter type checking is close to nonsense, the only stricter type check I'm aware of for our purposes is the requirement to cast void * to other types, the other types of strictness only apply when using C++ style casts.

RIght, C and C++ have different semantics, and while it might complain about some things, Perl is a C project, and any C++ warnings can't be taken literally because while the syntax is often the same, there's different rules about conversions and the rest of it.

Addresses Perl#21073 (but isn't a fix)

tonycoz · 2023-05-16T00:31:18Z

There are some additional constructs causing problems with MSVC C++ builds:

compound literals of the form (typename){ .field1= ..., ... } aren't available in C++. There's a construct* in C++ with a similar effect, but that syntax isn't C99 compatible: typename{ .field1 = ..., ... } (it has different semantics but I don't think they matter for us.
two cpan/ and one dist/ module guarded the perl headers with extern "C"{ ... }, which broke a window header that uses C++ constructs (templates) in C++ builds

There were some other more generic C++ issues.

Fixed by #21096, though I haven't updated version numbers (two modules are in cpan/)

* bearing in mind I haven't written C++ in a long, long time.

The second came up while trying to get perl to build as C++ in MSVC 2022. Addresses Perl#21073 (but isn't a fix)

clang-16 rejects this, MSVC warns about it. Part of fixing Perl#21073

These aren't needed with modern perls, and it's only useful to build modern perls as C++. This was causing an error in a system header when building blead as C++ with MSVC: C:\Program Files (x86)\Windows Kits\10\include\10.0.22000.0\um\wspiapi.h(53): error C2894: templates cannot be declared to have 'C' linkage See Perl/perl5#21073 for more information

We often build blead perl using a C++ compiler, but C++17 removed the register storage specifier, so some newer C++ compilers which default to C++17 now throw an error. See Perl/perl5#21073 for more discussion

These aren't needed with modern perls, and it's only useful to build modern perls as C++. This was causing an error in a system header when building blead as C++ with MSVC: C:\Program Files (x86)\Windows Kits\10\include\10.0.22000.0\um\wspiapi.h(53): error C2894: templates cannot be declared to have 'C' linkage See Perl/perl5#21073 for more information

clang-16 rejects this, MSVC warns about it. Part of fixing Perl#21073

clang-16 rejects this, MSVC warns about it. Part of fixing #21073

The second came up while trying to get perl to build as C++ in MSVC 2022. Addresses #21073 (but isn't a fix)

clang-16 rejects this, MSVC warns about it. Part of fixing Perl#21073

The second came up while trying to get perl to build as C++ in MSVC 2022. Addresses Perl#21073 (but isn't a fix)

tonycoz added the Needs Triage label May 4, 2023

Leont mentioned this issue May 8, 2023

Disambiguate array initialization designators in opcode.h #21081

Closed

jkeenan removed the Needs Triage label May 10, 2023

Leont mentioned this issue May 12, 2023

Revert usage of C99 array designators #21090

Merged

tonycoz added a commit to tonycoz/perl5 that referenced this issue May 14, 2023

perlhacktips: C99, but not array designators

5ac280c

Addresses Perl#21073 (but isn't a fix)

This was referenced May 14, 2023

perlhacktips: C99, but not array designators nor compound literals #21095

Merged

fix C++ build for MSVC 2022 #21096

Merged

tonycoz added a commit to tonycoz/perl5 that referenced this issue May 16, 2023

perlhacktips: C99, but not array designators nor compound literals

79a1755

The second came up while trying to get perl to build as C++ in MSVC 2022. Addresses Perl#21073 (but isn't a fix)

tonycoz added a commit to tonycoz/perl5 that referenced this issue May 18, 2023

C++ compatibility: eliminate "register" storage class usage

844964a

clang-16 rejects this, MSVC warns about it. Part of fixing Perl#21073

tonycoz mentioned this issue May 22, 2023

MD5.xs: eliminate C++ guards Dual-Life/digest-md5#24

Merged

tonycoz mentioned this issue May 22, 2023

eliminate use of register keyword for C++ compatibility Dual-Life/mime-base64#17

Open

tonycoz mentioned this issue May 22, 2023

Piece.xs: eliminate C++ guards Dual-Life/Time-Piece#64

Open

tonycoz added a commit to tonycoz/perl5 that referenced this issue May 22, 2023

C++ compatibility: eliminate "register" storage class usage

cbf6d0f

clang-16 rejects this, MSVC warns about it. Part of fixing Perl#21073

Leont closed this as completed in #21096 May 24, 2023

Leont pushed a commit that referenced this issue May 24, 2023

C++ compatibility: eliminate "register" storage class usage

492d6b6

clang-16 rejects this, MSVC warns about it. Part of fixing #21073

tonycoz added a commit that referenced this issue Jul 3, 2023

perlhacktips: C99, but not array designators nor compound literals

dcc4b62

The second came up while trying to get perl to build as C++ in MSVC 2022. Addresses #21073 (but isn't a fix)

khwilliamson pushed a commit to khwilliamson/perl5 that referenced this issue Jul 10, 2023

C++ compatibility: eliminate "register" storage class usage

c8167ac

clang-16 rejects this, MSVC warns about it. Part of fixing Perl#21073

khwilliamson pushed a commit to khwilliamson/perl5 that referenced this issue Jul 10, 2023

perlhacktips: C99, but not array designators nor compound literals

d2ededa

The second came up while trying to get perl to build as C++ in MSVC 2022. Addresses Perl#21073 (but isn't a fix)

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

C++ builds fail with modern C++ compilers #21073

C++ builds fail with modern C++ compilers #21073

tonycoz commented May 4, 2023

tonycoz commented May 4, 2023

khwilliamson commented May 7, 2023

tonycoz commented May 8, 2023

Leont commented May 8, 2023 •

edited

Loading

khwilliamson commented May 11, 2023

Leont commented May 12, 2023

leonerd commented May 12, 2023

leonerd commented May 12, 2023

thesamesam commented May 13, 2023

tonycoz commented May 16, 2023

C++ builds fail with modern C++ compilers #21073

C++ builds fail with modern C++ compilers #21073

Comments

tonycoz commented May 4, 2023

tonycoz commented May 4, 2023

khwilliamson commented May 7, 2023

tonycoz commented May 8, 2023

Leont commented May 8, 2023 • edited Loading

khwilliamson commented May 11, 2023

Leont commented May 12, 2023

leonerd commented May 12, 2023

leonerd commented May 12, 2023

thesamesam commented May 13, 2023

tonycoz commented May 16, 2023

Leont commented May 8, 2023 •

edited

Loading