bug#35650: [PATCH] Improving HP-UX support

[Pádraig Brady]

On 09/05/19 03:50, Osipov, Michael wrote:
> Hi folks,
> 
> please find a patch attached to properly compile and link GNU coreutils 
> 8.31 on HP-UX IA64. As of now, it does not compile out of the box.
> 
> The patch includes these changes:
> 
> * basenc.c: HP aCC does not allow anonymous unions inside structs, I 
> have named it otherwise the compilation fails
> * blake2/blake2.h: __attribute__((packed)) does not exist in HP aCC, use 
> #pragma pack
> * copy.c, mkfifo.c: disable SELinux code if it is not available
> * system.h: use __attribute(x) on a GNU-style compiler only
> * configure.ac, local.mk: use HP aCC-style linker options for libstdbuf
> 
> Compile instructions:
>> export PREFIX=/opt/ports/coreutils
>> export LIBDIR=$PREFIX/lib/hpux32
>> export CONFIGURE="./configure --prefix=$PREFIX --libdir=$LIBDIR"
>> export CPPFLAGS="-I$PREFIX/include -D_INCLUDE_STDC__SOURCE_199901"
>> export LDFLAGS="-L$LIBDIR"
>> autoreconf -fi
>> FORCE_UNSAFE_CONFIGURE=1 gl_cv_have_include_next=no ac_cv_func_getacl=no 
>> ac_cv_func_aclsort=no ac_cv_header_sys_bitypes_h=no $CONFIGURE
>> gmake install
> The patch isn't perfect, some issues need to be discussion because I 
> don't know yet how to solve them:
> * copy.c, mkfifo.c: for some reason the code compiles w/o changes 
> although SELinux is not available

These selinux interfaces are provided by gnulib,
and so should automatically by ignored on HPUX.
I.E. I'm not sure this part is needed at all.

> * I have disabled the compilation of b2sum completely because gmake 
> repeatedly quits with:
>> gmake[2]: *** No rule to make target '/var/tmp/AAA000546.i', needed by 
>> 'src/b2sum-md5sum.o'.  Stop.
>> gmake[2]: Leaving directory '/tmp/system-compile/gnu/coreutils-8.31'
>> gmake[1]: *** [Makefile:12647: all-recursive] Error 1
>> gmake[1]: Leaving directory '/tmp/system-compile/gnu/coreutils-8.31'
>> gmake: *** [Makefile:6827: all] Error 2


> diff -ur src/blake2/blake2.h src/blake2/blake2.h
> --- src/blake2/blake2.h       2018-05-14 06:20:24 +0000
> +++ src/blake2/blake2.h       2019-05-08 15:08:42 +0000
> @@ -21,8 +21,12 @@
>  #if defined(_MSC_VER)
>  #define BLAKE2_PACKED(x) __pragma(pack(push, 1)) x __pragma(pack(pop))
>  #else
> +#ifdef __hpux
> +#define BLAKE2_PACKED(x) x
>
> +#else
>  #define BLAKE2_PACKED(x) x __attribute__((packed))
>  #endif
> +#endif
>  
>  #if defined(__cplusplus)
>  extern "C" {
> @@ -86,6 +90,9 @@
>      size_t        outlen;
>    } blake2bp_state;
>  
> +#ifdef __hpux
> +#pragma pack 1
> +#endif
>  
>    BLAKE2_PACKED(struct blake2s_param__
>    {
> @@ -102,9 +109,15 @@
>      uint8_t  salt[BLAKE2S_SALTBYTES]; /* 24 */
>      uint8_t  personal[BLAKE2S_PERSONALBYTES];  /* 32 */
>    });
> +#ifdef __hpux
> +#pragma pack
> +#endif

That's a bit awkward,
also this file comes from blake2 upstream source on github,
so you might want to broach the change there.

> diff -ur src/system.h src/system.h
> --- src/system.h      2019-01-05 11:36:22 +0000
> +++ src/system.h      2019-05-08 15:19:29 +0000
> @@ -422,6 +422,10 @@
>  # endif
>  #endif
>  
> +#ifndef __GNUC__
> +#define __attribute(x) /* empty */
> +#endif

The only use of __attribute is:

src/libstdbuf.c:/* Use __attribute to avoid elision of __attribute__ on 
SUNPRO_C etc.  */
src/libstdbuf.c:static void __attribute ((constructor))

So we should probably have special __hpux handling in that file
rather than doing the above.

This is all dependent on stdbuf actually working on your platform.
Possibly the most appropriate change would be to use __attribute
rather than __attribute__ in configure.ac

>  #ifndef ATTRIBUTE_NORETURN
>  # define ATTRIBUTE_NORETURN __attribute__ ((__noreturn__))
>  #endif
> --- configure.ac      2019-03-04 08:40:55 +0000
> +++ configure.ac      2019-05-08 15:40:41 +0000
> @@ -481,8 +481,8 @@
>  gl_WARN_ADD([-errwarn], [CFLAGS])
>  # Put this message here, after gl_WARN_ADD's chatter.
>  AC_MSG_CHECKING([whether this system supports stdbuf])
> -CFLAGS="-fPIC $CFLAGS"
> -LDFLAGS="-shared $LDFLAGS"
> +CFLAGS="+z $CFLAGS"
> +LDFLAGS="-b $LDFLAGS"

This would need to be conditional,
again if stdbuf does actually work on hpux.

thanks,
Pádraig