Обсуждение: Solaris 9 and PgSQL 8.1.3 - createdb Link Fails
I have been unable to get PostreSQL 8.1.3 to compile on Solaris 9.
Thinking this might be a problem with the compile, I have tried GCC
3.3, GCC 3.4.3, and Sun Workshop Compiler 5.0. All three error while
trying to link the createdb binary. I'm configuring GCC builds using:
./configure --prefix=/usr/local/postgres-8.1.3 --without-readline
and Sun Workshop builds using:
./configure CC=/opt/SUNWspro/bin/cc --prefix=/usr/local/
postgres-8.1.3 --without-readline
A GCC build yields the following errors....
==========
gmake[4]: Entering directory `/home/swaltner/postgresql-8.1.3/src/
backend/parser'
gmake[4]: `keywords.o' is up to date.
gmake[4]: Leaving directory `/home/swaltner/postgresql-8.1.3/src/
backend/parser'
gcc -O2 -Wall -Wmissing-prototypes -Wpointer-arith -Winline -
Wdeclaration-after-statement -Wendif-labels -fno-strict-aliasing -
DFRONTEND -I../../../src/bin/pg_dump -I../../../src/bin/psql -
I../../../src/interfaces/libpq -I../../../src/include -c -o
createdb.o createdb.c
gcc -O2 -Wall -Wmissing-prototypes -Wpointer-arith -Winline -
Wdeclaration-after-statement -Wendif-labels -fno-strict-aliasing -
DFRONTEND -I../../../src/bin/pg_dump -I../../../src/bin/psql -
I../../../src/interfaces/libpq -I../../../src/include -c -o
common.o common.c
rm -f dumputils.c && ln -s ../../../src/bin/pg_dump/dumputils.c .
gcc -O2 -Wall -Wmissing-prototypes -Wpointer-arith -Winline -
Wdeclaration-after-statement -Wendif-labels -fno-strict-aliasing -
DFRONTEND -I../../../src/bin/pg_dump -I../../../src/bin/psql -
I../../../src/interfaces/libpq -I../../../src/include -c -o
dumputils.o dumputils.c
gcc -O2 -Wall -Wmissing-prototypes -Wpointer-arith -Winline -
Wdeclaration-after-statement -Wendif-labels -fno-strict-aliasing -
DFRONTEND -I../../../src/bin/pg_dump -I../../../src/bin/psql -
I../../../src/interfaces/libpq -I../../../src/include -L../../../src/
port -Wl,-R/usr/local/postgres-8.1.3/lib createdb.c createdb.o
common.o dumputils.o ../../../src/backend/parser/keywords.o -o
createdb
ld: fatal: symbol `main' is multiply-defined:
(file /tmp/ccZR65J8.o type=FUNC; file createdb.o type=FUNC);
ld: fatal: File processing errors. No output written to createdb
collect2: ld returned 1 exit status
gmake[3]: *** [createdb] Error 1
gmake[3]: Leaving directory `/home/swaltner/postgresql-8.1.3/src/bin/
scripts'
gmake[2]: *** [all] Error 2
gmake[2]: Leaving directory `/home/swaltner/postgresql-8.1.3/src/bin'
gmake[1]: *** [all] Error 2
gmake[1]: Leaving directory `/home/swaltner/postgresql-8.1.3/src'
gmake: *** [all] Error 2
hyperion:~/postgresql-8.1.3>
==========
Sun Workshop builds yield the following:
==========
gmake[4]: Entering directory `/home/swaltner/postgresql-8.1.3/src/
backend/parser'
gmake[4]: `keywords.o' is up to date.
gmake[4]: Leaving directory `/home/swaltner/postgresql-8.1.3/src/
backend/parser'
/opt/SUNWspro/bin/cc -Xa -v -DSUNOS4_CC -O -DFRONTEND -I../../../src/
bin/pg_dump -I../../../src/bin/psql -I../../../src/interfaces/libpq -
I../../../src/include -c -o createdb.o createdb.c
"createdb.c", line 212: warning: Function has no return statement : main
/opt/SUNWspro/bin/cc -Xa -v -DSUNOS4_CC -O -DFRONTEND -I../../../src/
bin/pg_dump -I../../../src/bin/psql -I../../../src/interfaces/libpq -
I../../../src/include -c -o common.o common.c
rm -f dumputils.c && ln -s ../../../src/bin/pg_dump/dumputils.c .
/opt/SUNWspro/bin/cc -Xa -v -DSUNOS4_CC -O -DFRONTEND -I../../../src/
bin/pg_dump -I../../../src/bin/psql -I../../../src/interfaces/libpq -
I../../../src/include -c -o dumputils.o dumputils.c
/opt/SUNWspro/bin/cc -Xa -v -DSUNOS4_CC -O -DFRONTEND -I../../../src/
bin/pg_dump -I../../../src/bin/psql -I../../../src/interfaces/libpq -
I../../../src/include -L../../../src/port -Wl,-R/usr/local/
postgres-8.1.3/lib createdb.c createdb.o common.o
dumputils.o ../../../src/backend/parser/keywords.o -o createdb
"createdb.c", line 212: warning: Function has no return statement : main
Undefined first referenced
symbol in file
appendPQExpBuffer createdb.o
PQresultStatus createdb.o
PQerrorMessage createdb.o
initPQExpBuffer createdb.o
destroyPQExpBuffer dumputils.o
PQexec createdb.o
PQclear createdb.o
pg_char_to_encoding createdb.o
appendPQExpBufferChar dumputils.o
PQfinish createdb.o
resetPQExpBuffer dumputils.o
getopt_long createdb.o
PQstatus common.o
set_pglocale_pgservice createdb.o
simple_prompt common.o
createPQExpBuffer dumputils.o
printfPQExpBuffer createdb.o
appendPQExpBufferStr dumputils.o
PQsetdbLogin common.o
get_progname createdb.o
ld: fatal: Symbol referencing errors. No output written to createdb
gmake[3]: *** [createdb] Error 1
gmake[3]: Leaving directory `/home/swaltner/postgresql-8.1.3/src/bin/
scripts'
gmake[2]: *** [all] Error 2
gmake[2]: Leaving directory `/home/swaltner/postgresql-8.1.3/src/bin'
gmake[1]: *** [all] Error 2
gmake[1]: Leaving directory `/home/swaltner/postgresql-8.1.3/src'
gmake: *** [all] Error 2
252.93u 44.02s 6:09.17 80.4%
hyperion:~/postgresql-8.1.3>
==========
I've done several searches to try and find info about this, but the
only hits on createdb that I find are related to using it, not
getting it to build. I'll try downloading an older release of
PostgreSQL to see if that changes the behavior. Any suggestions would
be greatly appreciated.
Not on this system. I successfully compiled version 7.2.1 on a different Solaris 9 system back in Nov 2004. I could always copy that installation across. I receive the same "main multiply-defined" error when trying to compile PostgreSQL 8.0.7 and 8.1.2 on this same system. I'm trying 7.4.12 now and will also install readline to see if the "./configure --without-readline" option is giving me trouble. Steve On Feb 22, 2006, at 11:33 AM, Mark Round wrote: > Do you have an older version of PostgreSQL installed ? > > On 2/22/06, Steve Waltner <steve.waltner@engenio.com> wrote: >> I have been unable to get PostreSQL 8.1.3 to compile on Solaris 9.
I tried the following: - Compile PgSQL versions 8.1.2, 8.0.7, 7.4.12 on this new system and another similar system - Install readline 5.1 and repeat the above - "gmake -j 4" on 8 CPU system - Compile PgSQL 8.1.3 on the original system I loaded PgSQL onto The first three failed, but the last completed successfully. Looking further into the failure, I noticed that the build that completed was running an old version of GNU make (3.79.1), while the failures occurred with GNU make 3.81beta2. We installed the GNU make 3.81beta2 http://make.paulandlesley.org/make-3.81beta2.tar.bz2 to solve the problem of getting "interrupted system call" errors when running parallel builds with the source code stored on a NFS server. I haven't seen any other software packages fail to build when using this GNU make. There might be something screwy in the PostgreSQL makefiles though since I did notice the following behavior. GNU make 3.81beta2 initial build ========================== [snip] gcc -O2 -Wall -Wmissing-prototypes -Wpointer-arith -Winline -Wendif- labels -fno-strict-aliasing -DFRONTEND -I../../../src/bin/pg_dump - I../../../src/bin/psql -I../../../src/interfaces/libpq -I../../../src/ include -I/soft/gnu/readline/5.1/include -c -o createdb.o createdb.c [snip] gcc -O2 -Wall -Wmissing-prototypes -Wpointer-arith -Winline -Wendif- labels -fno-strict-aliasing -DFRONTEND -I../../../src/bin/pg_dump - I../../../src/bin/psql -I../../../src/interfaces/libpq -I../../../src/ include -I/soft/gnu/readline/5.1/include -L../../../src/port -Wl,-R/ usr/local/postgresql-8.1.3/lib createdb.c createdb.o common.o dumputils.o ../../../src/backend/parser/keywords.o -o createdb ld: fatal: symbol `main' is multiply-defined: (file /tmp/ccIwnYOH.o type=FUNC; file createdb.o type=FUNC); ld: fatal: File processing errors. No output written to createdb ========================== GNU make 3.81beta2 second build (ie: don't run "gmake clean") ========================== [snip] gcc -O2 -Wall -Wmissing-prototypes -Wpointer-arith -Winline -Wendif- labels -fno-strict-aliasing createdb.o common.o dumputils.o ../../../ src/backend/parser/keywords.o -L../../../src/port -lpgport -L../../../ src/interfaces/libpq -lpq -L../../../src/port -Wl,-R/usr/local/ postgresql-8.1.3/lib -lpgport -lz -lreadline -ltermcap -lrt -lresolv - lgen -lsocket -lnsl -ldl -lm -o createdb [snip] ========================== GNU make 3.80 initial build ========================== [snip] gcc -O2 -Wall -Wmissing-prototypes -Wpointer-arith -Winline -Wendif- labels -fno-strict-aliasing -DFRONTEND -I../../../src/bin/pg_dump - I../../../src/bin/psql -I../../../src/interfaces/libpq -I../../../src/ include -I/soft/gnu/readline/5.1/include -c -o createdb.o createdb.c [snip] gcc -O2 -Wall -Wmissing-prototypes -Wpointer-arith -Winline -Wendif- labels -fno-strict-aliasing createdb.o common.o dumputils.o ../../../ src/backend/parser/keywords.o -L../../../src/port -lpgport -L../../../ src/interfaces/libpq -lpq -L../../../src/port -Wl,-R/usr/local/ postgresql-8.1.3/lib -lpgport -lz -lreadline -ltermcap -lrt -lresolv - lgen -lsocket -lnsl -ldl -lm -o createdb [snip] ========================== This problem is not isolated to the createdb command. It happened on every single binary in .../src/bin/scripts/. I needed to run the build 10 times to make it through this directory when using the latest version of GNU make. Note that the initial build of the files in .../src/bin/scripts/ that fail all have greatly different syntax on the link command than the ones that link properly. When the make is run using the released version of GNU make or a when the link is reattempted with 3.81beta2, the link arguments are much different. I've gotten the software compile, but wanted to make sure people were aware of this issue. Steve On Feb 22, 2006, at 11:52 AM, Steve Waltner wrote: > Not on this system. > > I successfully compiled version 7.2.1 on a different Solaris 9 > system back in Nov 2004. I could always copy that installation across. > > I receive the same "main multiply-defined" error when trying to > compile PostgreSQL 8.0.7 and 8.1.2 on this same system. I'm trying > 7.4.12 now and will also install readline to see if the "./ > configure --without-readline" option is giving me trouble. > > Steve > > On Feb 22, 2006, at 11:33 AM, Mark Round wrote: >> Do you have an older version of PostgreSQL installed ? >> >> On 2/22/06, Steve Waltner <steve.waltner@engenio.com> wrote: >>> I have been unable to get PostreSQL 8.1.3 to compile on Solaris 9.
Steve Waltner wrote: > occurred with GNU make 3.81beta2. We installed the GNU make 3.81beta2 > http://make.paulandlesley.org/make-3.81beta2.tar.bz2 to solve the I can reproduce these failures with that make version, but considering that everything before and after that version works, there will be little interest in debugging this. -- Peter Eisentraut http://developer.postgresql.org/~petere/
Do you have an older version of PostgreSQL installed ? On 2/22/06, Steve Waltner <steve.waltner@engenio.com> wrote: > I have been unable to get PostreSQL 8.1.3 to compile on Solaris 9.
The beta2 release is many, many months old. We got up to beta4, and are
now testing rc1.
Please try the latest pretest:
ftp://alpha.gnu.org/gnu/make/make-3.81rc1.tar.gz
If you still have problems NOW is the time to say something, since
unless we discover show-stoppers I'm going to release 3.81 this coming
weekend.
--
-------------------------------------------------------------------------------
Paul D. Smith <psmith@gnu.org> Find some GNU make tips at:
http://www.gnu.org http://make.paulandlesley.org
"Please remain calm...I may be mad, but I am a professional." --Mad Scientist
Paul, Thanks for the link to the new make source code. I hadn't originally found the beta source code on alpha.gnu.org when I downloaded the 3.81beta2 code back in September, so I didn't think to look there. I did check ftp.gnu.org and found out 3.80 was still the official release. Upgrading GNU make did the trick. I'm able to compile the PostgreSQL source code using a "make -j 8" command on our Sun V490. This compilation problem appears to have simply been a bug in GNU Make 3.81beta2 that was fixed somewhere along the way to 3.81rc1. Thanks again for the help. Steve On Feb 22, 2006, at 11:01 PM, Paul D. Smith wrote: > The beta2 release is many, many months old. We got up to beta4, > and are > now testing rc1. > > Please try the latest pretest: > > ftp://alpha.gnu.org/gnu/make/make-3.81rc1.tar.gz > > If you still have problems NOW is the time to say something, since > unless we discover show-stoppers I'm going to release 3.81 this coming > weekend.
%% Steve Waltner <swaltner@engenio.com> writes: sw> Thanks for the link to the new make source code. I hadn't sw> originally found the beta source code on alpha.gnu.org when I sw> downloaded the 3.81beta2 code back in September, so I didn't think sw> to look there. Yes; there was a problem with the secure uploads to the GNU ftp servers (both alpha and ftp) back then so I hosted that beta myself. That problem was later fixed, so beta3, beta4, and rc1 have all been hosted on alpha. sw> Upgrading GNU make did the trick. I'm able to compile the sw> PostgreSQL source code using a "make -j 8" command on our Sun sw> V490. This compilation problem appears to have simply been a bug sw> in GNU Make 3.81beta2 that was fixed somewhere along the way to sw> 3.81rc1. Good to know! Thanks for testing this. Cheers! -- ------------------------------------------------------------------------------- Paul D. Smith <psmith@gnu.org> Find some GNU make tips at: http://www.gnu.org http://make.paulandlesley.org "Please remain calm...I may be mad, but I am a professional." --Mad Scientist