Re: BUG #6758: ./configure script sets HAVE_WCSTOMBS_L 1

Andrew Hastie <andrew@ahastie.net> writes:

I'm currently working on a project where I need to get PGv9.1 up and
running on an IBM AIXv7.1 server, so I do have access to a suitable
machine for a period of time if I can provide any further diags to help
resolve the issue.

That would be helpful, because both of the previous complainants seem
to have wandered off without helping us get to the bottom of it :-(.

What I'd like to see at this point is the portion of config.log where
the configure script is trying to probe for wcstombs_l. For example,
on a machine that has wcstombs and not wcstombs_l, I see:

...
configure:19272: checking for wcstombs
configure:19328: gcc -o conftest -O2 -Wall -Wmissing-prototypes -Wpointer-arith -Wformat-security -fno-strict-aliasing -g -D_XOPEN_SOURCE_EXTENDED conftest.c -lz -lreadline -ltermcap -lm >&5
configure:19335: $? = 0
configure:19357: result: yes
configure:19272: checking for wcstombs_l
configure:19328: gcc -o conftest -O2 -Wall -Wmissing-prototypes -Wpointer-arith -Wformat-security -fno-strict-aliasing -g -D_XOPEN_SOURCE_EXTENDED conftest.c -lz -lreadline -ltermcap -lm >&5
/usr/ccs/bin/ld: Unsatisfied symbols:
wcstombs_l (code)
collect2: ld returned 1 exit status
configure:19335: $? = 1
configure: failed program was:
| /* confdefs.h. */
| #define PACKAGE_NAME "PostgreSQL"
| #define PACKAGE_TARNAME "postgresql"
| #define PACKAGE_VERSION "9.3devel"
| #define PACKAGE_STRING "PostgreSQL 9.3devel"
| #define PACKAGE_BUGREPORT "pgsql-bugs@postgresql.org"
| #define PG_VERSION "9.3devel"
| #define PG_MAJORVERSION "9.3"
| #define USE_INTEGER_DATETIMES 1
| #define DEF_PGPORT 5440
| #define DEF_PGPORT_STR "5440"
| #define BLCKSZ 8192
| #define RELSEG_SIZE 131072
| #define XLOG_BLCKSZ 8192
| #define XLOG_SEG_SIZE (16 * 1024 * 1024)
| #define USE_ASSERT_CHECKING 1
| #define PG_KRB_SRVNAM "postgres"
| #define HAVE_LIBM 1
| #define HAVE_LIBREADLINE 1
| #define HAVE_LIBZ 1
| #define HAVE_SPINLOCKS 1
| #define STDC_HEADERS 1
| #define HAVE_SYS_TYPES_H 1
| #define HAVE_SYS_STAT_H 1
| #define HAVE_STDLIB_H 1
| #define HAVE_STRING_H 1
| #define HAVE_MEMORY_H 1
| #define HAVE_STRINGS_H 1
| #define HAVE_INTTYPES_H 1
| #define HAVE_UNISTD_H 1
| #define HAVE_CRYPT_H 1
| #define HAVE_LANGINFO_H 1
| #define HAVE_POLL_H 1
| #define HAVE_PWD_H 1
| #define HAVE_SYS_IOCTL_H 1
| #define HAVE_SYS_IPC_H 1
| #define HAVE_SYS_POLL_H 1
| #define HAVE_SYS_PSTAT_H 1
| #define HAVE_SYS_RESOURCE_H 1
| #define HAVE_SYS_SEM_H 1
| #define HAVE_SYS_SHM_H 1
| #define HAVE_SYS_SOCKET_H 1
| #define HAVE_SYS_TIME_H 1
| #define HAVE_SYS_UN_H 1
| #define HAVE_TERMIOS_H 1
| #define HAVE_UTIME_H 1
| #define HAVE_WCHAR_H 1
| #define HAVE_NET_IF_H 1
| #define HAVE_NETINET_IN_H 1
| #define HAVE_NETINET_TCP_H 1
| #define HAVE_READLINE_READLINE_H 1
| #define HAVE_READLINE_HISTORY_H 1
| #define WORDS_BIGENDIAN 1
| #define USE_INLINE 1
| #define HAVE_STRINGIZE 1
| #define FLEXIBLE_ARRAY_MEMBER 1
| #define HAVE_FUNCNAME__FUNC 1
| #define HAVE_TZNAME 1
| #define HAVE_UNIX_SOCKETS 1
| #define HAVE_INTPTR_T 1
| #define HAVE_UINTPTR_T 1
| #define HAVE_LONG_LONG_INT 1
| #define _FILE_OFFSET_BITS 64
| #define SIZEOF_OFF_T 8
| #define HAVE_INT_TIMEZONE /**/
| #define ACCEPT_TYPE_RETURN int
| #define ACCEPT_TYPE_ARG1 int
| #define ACCEPT_TYPE_ARG2 struct sockaddr *
| #define ACCEPT_TYPE_ARG3 size_t
| #define HAVE_CBRT 1
| #define HAVE_FDATASYNC 1
| #define HAVE_GETRLIMIT 1
| #define HAVE_MEMMOVE 1
| #define HAVE_POLL 1
| #define HAVE_PSTAT 1
| #define HAVE_READLINK 1
| #define HAVE_SETSID 1
| #define HAVE_SIGPROCMASK 1
| #define HAVE_SYMLINK 1
| #define HAVE_TOWLOWER 1
| #define HAVE_UTIME 1
| #define HAVE_UTIMES 1
| #define HAVE_WCSTOMBS 1
| /* end confdefs.h. */
| /* Define wcstombs_l to an innocuous variant, in case <limits.h> declares wcstombs_l.
| For example, HP-UX 11i <limits.h> declares gettimeofday. */
| #define wcstombs_l innocuous_wcstombs_l
|
| /* System header to define __stub macros and hopefully few prototypes,
| which can conflict with char wcstombs_l (); below.
| Prefer <limits.h> to <assert.h> if __STDC__ is defined, since
| <limits.h> exists even on freestanding compilers. */
|
| #ifdef __STDC__
| # include <limits.h>
| #else
| # include <assert.h>
| #endif
|
| #undef wcstombs_l
|
| /* Override any GCC internal prototype to avoid an error.
| Use char because int might match the return type of a GCC
| builtin and then its argument prototype would still apply. */
| #ifdef __cplusplus
| extern "C"
| #endif
| char wcstombs_l ();
| /* The GNU C library defines this for functions which it implements
| to always fail with ENOSYS. Some functions are actually named
| something starting with __ and the normal name is an alias. */
| #if defined __stub_wcstombs_l || defined __stub___wcstombs_l
| choke me
| #endif
|
| int
| main ()
| {
| return wcstombs_l ();
| ;
| return 0;
| }
configure:19357: result: no
configure:19374: checking for fseeko
...

It would also be interesting to check what would happen if configure
actually tried to probe for mbstowcs_l, instead of making assumptions.
To do that, copy the test program above (removing the "| " indentation),
replace "wcstombs_l" by "mbstowcs_l", and then try to compile it ---
but be sure to use the compile command shown in your config.log, not
mine.

regards, tom lane

On 29/08/12 18:16, Tom Lane wrote:

Andrew Hastie <andrew@ahastie.net> writes:

I'm currently working on a project where I need to get PGv9.1 up and
running on an IBM AIXv7.1 server, so I do have access to a suitable
machine for a period of time if I can provide any further diags to help
resolve the issue.

That would be helpful, because both of the previous complainants seem
to have wandered off without helping us get to the bottom of it :-(.

What I'd like to see at this point is the portion of config.log where
the configure script is trying to probe for wcstombs_l. For example,
on a machine that has wcstombs and not wcstombs_l, I see:

...
configure:19272: checking for wcstombs
configure:19328: gcc -o conftest -O2 -Wall -Wmissing-prototypes -Wpointer-arith -Wformat-security -fno-strict-aliasing -g -D_XOPEN_SOURCE_EXTENDED conftest.c -lz -lreadline -ltermcap -lm >&5
configure:19335: $? = 0
configure:19357: result: yes
configure:19272: checking for wcstombs_l
configure:19328: gcc -o conftest -O2 -Wall -Wmissing-prototypes -Wpointer-arith -Wformat-security -fno-strict-aliasing -g -D_XOPEN_SOURCE_EXTENDED conftest.c -lz -lreadline -ltermcap -lm >&5
/usr/ccs/bin/ld: Unsatisfied symbols:
wcstombs_l (code)
collect2: ld returned 1 exit status
configure:19335: $? = 1
configure: failed program was:
| /* confdefs.h. */
| #define PACKAGE_NAME "PostgreSQL"
| #define PACKAGE_TARNAME "postgresql"
| #define PACKAGE_VERSION "9.3devel"
| #define PACKAGE_STRING "PostgreSQL 9.3devel"
| #define PACKAGE_BUGREPORT "pgsql-bugs@postgresql.org"
| #define PG_VERSION "9.3devel"
| #define PG_MAJORVERSION "9.3"
| #define USE_INTEGER_DATETIMES 1
| #define DEF_PGPORT 5440
| #define DEF_PGPORT_STR "5440"
| #define BLCKSZ 8192
| #define RELSEG_SIZE 131072
| #define XLOG_BLCKSZ 8192
| #define XLOG_SEG_SIZE (16 * 1024 * 1024)
| #define USE_ASSERT_CHECKING 1
| #define PG_KRB_SRVNAM "postgres"
| #define HAVE_LIBM 1
| #define HAVE_LIBREADLINE 1
| #define HAVE_LIBZ 1
| #define HAVE_SPINLOCKS 1
| #define STDC_HEADERS 1
| #define HAVE_SYS_TYPES_H 1
| #define HAVE_SYS_STAT_H 1
| #define HAVE_STDLIB_H 1
| #define HAVE_STRING_H 1
| #define HAVE_MEMORY_H 1
| #define HAVE_STRINGS_H 1
| #define HAVE_INTTYPES_H 1
| #define HAVE_UNISTD_H 1
| #define HAVE_CRYPT_H 1
| #define HAVE_LANGINFO_H 1
| #define HAVE_POLL_H 1
| #define HAVE_PWD_H 1
| #define HAVE_SYS_IOCTL_H 1
| #define HAVE_SYS_IPC_H 1
| #define HAVE_SYS_POLL_H 1
| #define HAVE_SYS_PSTAT_H 1
| #define HAVE_SYS_RESOURCE_H 1
| #define HAVE_SYS_SEM_H 1
| #define HAVE_SYS_SHM_H 1
| #define HAVE_SYS_SOCKET_H 1
| #define HAVE_SYS_TIME_H 1
| #define HAVE_SYS_UN_H 1
| #define HAVE_TERMIOS_H 1
| #define HAVE_UTIME_H 1
| #define HAVE_WCHAR_H 1
| #define HAVE_NET_IF_H 1
| #define HAVE_NETINET_IN_H 1
| #define HAVE_NETINET_TCP_H 1
| #define HAVE_READLINE_READLINE_H 1
| #define HAVE_READLINE_HISTORY_H 1
| #define WORDS_BIGENDIAN 1
| #define USE_INLINE 1
| #define HAVE_STRINGIZE 1
| #define FLEXIBLE_ARRAY_MEMBER 1
| #define HAVE_FUNCNAME__FUNC 1
| #define HAVE_TZNAME 1
| #define HAVE_UNIX_SOCKETS 1
| #define HAVE_INTPTR_T 1
| #define HAVE_UINTPTR_T 1
| #define HAVE_LONG_LONG_INT 1
| #define _FILE_OFFSET_BITS 64
| #define SIZEOF_OFF_T 8
| #define HAVE_INT_TIMEZONE /**/
| #define ACCEPT_TYPE_RETURN int
| #define ACCEPT_TYPE_ARG1 int
| #define ACCEPT_TYPE_ARG2 struct sockaddr *
| #define ACCEPT_TYPE_ARG3 size_t
| #define HAVE_CBRT 1
| #define HAVE_FDATASYNC 1
| #define HAVE_GETRLIMIT 1
| #define HAVE_MEMMOVE 1
| #define HAVE_POLL 1
| #define HAVE_PSTAT 1
| #define HAVE_READLINK 1
| #define HAVE_SETSID 1
| #define HAVE_SIGPROCMASK 1
| #define HAVE_SYMLINK 1
| #define HAVE_TOWLOWER 1
| #define HAVE_UTIME 1
| #define HAVE_UTIMES 1
| #define HAVE_WCSTOMBS 1
| /* end confdefs.h. */
| /* Define wcstombs_l to an innocuous variant, in case <limits.h> declares wcstombs_l.
| For example, HP-UX 11i <limits.h> declares gettimeofday. */
| #define wcstombs_l innocuous_wcstombs_l
|
| /* System header to define __stub macros and hopefully few prototypes,
| which can conflict with char wcstombs_l (); below.
| Prefer <limits.h> to <assert.h> if __STDC__ is defined, since
| <limits.h> exists even on freestanding compilers. */
|
| #ifdef __STDC__
| # include <limits.h>
| #else
| # include <assert.h>
| #endif
|
| #undef wcstombs_l
|
| /* Override any GCC internal prototype to avoid an error.
| Use char because int might match the return type of a GCC
| builtin and then its argument prototype would still apply. */
| #ifdef __cplusplus
| extern "C"
| #endif
| char wcstombs_l ();
| /* The GNU C library defines this for functions which it implements
| to always fail with ENOSYS. Some functions are actually named
| something starting with __ and the normal name is an alias. */
| #if defined __stub_wcstombs_l || defined __stub___wcstombs_l
| choke me
| #endif
|
| int
| main ()
| {
| return wcstombs_l ();
| ;
| return 0;
| }
configure:19357: result: no
configure:19374: checking for fseeko
...

It would also be interesting to check what would happen if configure
actually tried to probe for mbstowcs_l, instead of making assumptions.
To do that, copy the test program above (removing the "| " indentation),
replace "wcstombs_l" by "mbstowcs_l", and then try to compile it ---
but be sure to use the compile command shown in your config.log, not
mine.

regards, tom lane

Hi Tom,

Thanks for the pointers. Hopefully some of the following may shed some
light on the issue.

Here's the portion from config.log when probing for "wctsombs" and
"wctsombs_l" :-

configure:18948: checking for wcstombs
configure:19004: xlc -o conftest -qlanglvl=extc89 -qnoansialias
-I/opt/freeware/include -I/home/ahastie/gnu/zlib-1.2.7
-I/home/ahastie/gnu/readline-6.2 -L/opt/freeware/lib conftest.c -lz
-lreadline -l
m -lssl -lcrypto -lz -lreadline -lcurses -lld -lmass -lm >&5
configure:19011: $? = 0
configure:19033: result: yes
configure:18948: checking for wcstombs_l
configure:19004: xlc -o conftest -qlanglvl=extc89 -qnoansialias
-I/opt/freeware/include -I/home/ahastie/gnu/zlib-1.2.7
-I/home/ahastie/gnu/readline-6.2 -L/opt/freeware/lib conftest.c -lz
-lreadline -l
m -lssl -lcrypto -lz -lreadline -lcurses -lld -lmass -lm >&5
configure:19011: $? = 0
configure:19033: result: yes

So both probe OK. Here's the fail from gmake just to confirm the error:-

ld: 0711-317 ERROR: Undefined symbol: .mbstowcs_l
ld: 0711-345 Use the -bloadmap or -bnoquiet option to obtain more
information.
gmake[2]: *** [postgres] Error 8
gmake[2]: Leaving directory `/home/ahastie/postgresql-9.1.5/src/backend'
gmake[1]: *** [all-backend-recurse] Error 2
gmake[1]: Leaving directory `/home/ahastie/postgresql-9.1.5/src'
gmake: *** [all-src-recurse] Error 2

Using confdefs.c as a template, I created 2 tests where I only changed
the method names contained in the final 14 lines. Here are the results:-

Test #1
=====
Source (end portion only)

....
char wcstombs_l ();
/* The GNU C library defines this for functions which it implements
to always fail with ENOSYS. Some functions are actually named
something starting with __ and the normal name is an alias. */
#if defined __stub_wcstombs_l |defined __stub___wcstombs_l
choke me
#endif

int
main ()
{
return wcstombs_l ();
;
return 0;
}$

Results #1
=======
$ xlc -o conftest -qlanglvl=extc89 -qnoansialias -I/opt/freeware/include
-I/home/ahastie/gnu/zlib-1.2.7 -I/home/ahastie/gnu/readline-6.2
-L/opt/freeware/lib pgtest.c -lz -lreadline -lm -lssl -lcrypto -lz
-lreadline -lcurses -lld -lmass -lm
$

Test #2
=====

Source (end portion only)

.......
char mbstowcs_l ();
/* The GNU C library defines this for functions which it implements
to always fail with ENOSYS. Some functions are actually named
something starting with __ and the normal name is an alias. */
#if defined __stub_wcstombs_l |defined __stub___wcstombs_l
choke me
#endif

int
main ()
{
return mbstowcs_l ();
;
return 0;
}$

Results #2
=======
$ xlc -o conftest -qlanglvl=extc89 -qnoansialias -I/opt/freeware/include
-I/home/ahastie/gnu/zlib-1.2.7 -I/home/ahastie/gnu/readline-6.2
-L/opt/freeware/lib pgtest2.c -lz -lreadline -lm -lssl -lcrypto -lz
-lreadline -lcurses -lld -lmass -lm
ld: 0711-317 ERROR: Undefined symbol: .mbstowcs_l
ld: 0711-345 Use the -bloadmap or -bnoquiet option to obtain more
information.
$

If it helps, I tried the above call with the loadmap logging switched
on, and here is the output:-

(ld): halt 4
(ld): setfflag 4
(ld): savename conftest
(ld): filelist 14 1
(ld): i /lib/crt0.o
(ld): i pgtest2.o
(ld): lib /opt/freeware/lib/libz.a
(ld): lib /opt/freeware/lib/libreadline.a
(ld): lib /usr/lib/libm.a
(ld): lib /usr/lib/libssl.a
(ld): lib /usr/lib/libcrypto.a
(ld): lib /usr/lib/libcurses.a
(ld): lib /usr/lib/libld.a
(ld): lib /usr/lib/libmass.a
(ld): lib /usr/vac/lib/libxlopt.a
(ld): lib /usr/vac/lib/libxlipa.a
(ld): lib /usr/vac/lib/libxl.a
(ld): lib /usr/lib/libc.a
LIBRARY: Shared object libz.a[libz.so.1]: 72 symbols imported.
LIBRARY: Shared object libreadline.a[libreadline.so.4]: 398 symbols
imported.
LIBRARY: Shared object libssl.a[libssl.so.0.9.8]: 459 symbols imported.
LIBRARY: Shared object libcrypto.a[libcrypto.so.0.9.8]: 3040 symbols
imported.
LIBRARY: Shared object libcurses.a[shr42.o]: 508 symbols imported.
LIBRARY: Shared object libc.a[shr.o]: 3244 symbols imported.
LIBRARY: Shared object libc.a[meth.o]: 2 symbols imported.
LIBRARY: Shared object libc.a[posix_aio.o]: 20 symbols imported.
LIBRARY: Shared object libc.a[aio.o]: 18 symbols imported.
LIBRARY: Shared object libc.a[pse.o]: 8 symbols imported.
LIBRARY: Shared object libc.a[dl.o]: 4 symbols imported.
LIBRARY: Shared object libc.a[pty.o]: 1 symbols imported.
FILELIST: Number of previously inserted files processed: 14
(ld): resolve
RESOLVE: 30 of 13260 symbols were kept.
(ld): addgl /usr/lib/glink.o
ADDGL: Glink code added for 2 symbols.
(ld): er full
ld: 0711-318 ERROR: Undefined symbols were found.
The following symbols are in error:
Symbol Inpndx TY CL Source-File(Object-File) OR
Import-File{Shared-object}
RLD: Address Section Rld-type
Referencing Symbol
----------------------------------------------------------------------------------------------
.mbstowcs_l [20] ER PR pgtest2.c(pgtest2.o)
0000000c .text R_RBR [12] .main
ER: The return code is 8.

Again, let me know if you'd like me to test anything further.

Thanks,
Andrew

Andrew Hastie <andrew@ahastie.net> writes:

Thanks for the pointers. Hopefully some of the following may shed some
light on the issue.

Thanks for the results. It seems difficult to come to any conclusion
other than that AIX has got wcstombs_l but not mbstowcs_l ... which is
darn bizarre; I'd suggest reporting it as a bug to IBM.

In the meantime, it seems like what we had better do is explicitly
test for both functions, rather than making the assumption that they
would both or neither be present. I'm not sure if we can provide any
extra functionality if only wcstombs_l is available --- that will
have to emerge in making the patch. We can certainly build without
it though.

As a temporary workaround, you should probably just modify pg_config.h
after configure runs, to remove the HAVE_WCSTOMBS_L symbol.

regards, tom lane

On 31/08/12 15:59, Tom Lane wrote:

Andrew Hastie <andrew@ahastie.net> writes:

Thanks for the pointers. Hopefully some of the following may shed some
light on the issue.

Thanks for the results. It seems difficult to come to any conclusion
other than that AIX has got wcstombs_l but not mbstowcs_l ... which is
darn bizarre; I'd suggest reporting it as a bug to IBM.

In the meantime, it seems like what we had better do is explicitly
test for both functions, rather than making the assumption that they
would both or neither be present. I'm not sure if we can provide any
extra functionality if only wcstombs_l is available --- that will
have to emerge in making the patch. We can certainly build without
it though.

As a temporary workaround, you should probably just modify pg_config.h
after configure runs, to remove the HAVE_WCSTOMBS_L symbol.

Confirm that the workaround works just fine. Now have PG 9.1.5 up and
running on AIX just fine.
Thanks.
Andrew

Andrew Hastie <andrew@ahastie.net> writes:

On 31/08/12 15:59, Tom Lane wrote:

As a temporary workaround, you should probably just modify pg_config.h
after configure runs, to remove the HAVE_WCSTOMBS_L symbol.

Confirm that the workaround works just fine. Now have PG 9.1.5 up and
running on AIX just fine.

Thanks for checking. If you have time, perhaps you'd like to apply the
committed patch:
http://git.postgresql.org/gitweb/?p=postgresql.git;a=commitdiff;h=97395185b85b786523ee41225b53bd84c98d34f4
and confirm that all is well when using that.

regards, tom lane

Apologies for delay in progressing this, but I only have access to an
AIX7.1 system periodically.

Just to confirm that source build of 9.2.1 release now builds clean
against AIXv7.1 with xlc v12.1 compiler.

Thanks,
Andrew

On 31/08/12 20:20, Tom Lane wrote:

Andrew Hastie<andrew@ahastie.net> writes:

On 31/08/12 15:59, Tom Lane wrote:

As a temporary workaround, you should probably just modify pg_config.h
after configure runs, to remove the HAVE_WCSTOMBS_L symbol.

Confirm that the workaround works just fine. Now have PG 9.1.5 up and
running on AIX just fine.

Thanks for checking. If you have time, perhaps you'd like to apply the
committed patch:
http://git.postgresql.org/gitweb/?p=postgresql.git;a=commitdiff;h=97395185b85b786523ee41225b53bd84c98d34f4
and confirm that all is well when using that.

regards, tom lane

--

/*Andrew Hastie*/*
Software Engineer
Rocket.AS Support and Development
Rocket Software
*Innovation Centre · Warwick Technology Park, Warwick CV34 6UW · UK
Tel: +44.1926.482545
Email: andrew.hastie@rocketsoftware.com
<mailto:andrew.hastie@rocketsoftware.com>
Web: www.rocketsoftware.com/as <http://www.rocketsoftware.com/as&gt;

Apologies for delay in progressing this, but I only have access to an
AIX7.1 system periodically.

Just to confirm that source build of 9.2.1 release now builds clean
against AIXv7.1 with xlc v12.1 compiler.

Thanks,
Andrew