Thread: Windows on SuSE? 7.4
SuSE 7.3, PostgreSQL cvshead (7.4) This is as far as I've gotten with 7.4. It is built and configured like my 7.3 installation on the same machine. I have built from CVS previously. And the production sources always builds very nice and clean. Nothing runs. gbd output is below. It is balking on loading a suspect library. I ordinarily would suspect some minor permission or configuration issue, but I cannot see one. Joe Conway suggested it might be a problem with the the windows configuration changes. My machine has WINE on it (windows) although I never use it. There really should be no reason for anything to be trying to link in a windows socket library. I don't have time to follow it further right now. But if someone else has a similar configuration and/or wine installed, it would be very helpful to compare 7.4 installations. I can provide config arguments and log if someone would find them useful. >gdb bin/postmaster GNU gdb 20010316 Copyright 2001 Free Software Foundation, Inc. GDB is free software, covered by the GNU General Public License, and you are welcome to change it and/or distribute copies of it under certain conditions. Type "show copying" to see the conditions. There is absolutely no warranty for GDB. Type "show warranty" for details. This GDB was configured as "i386-suse-linux"... (gdb) run -i -D /local/pghead/data Starting program: /local/pghead/bin/postmaster -i -D /local/pghead/data Program received signal SIGSEGV, Segmentation fault. 0x40099ac5 in dllname () from /usr/lib/libwsock32.so (gdb) bt #0 0x40099ac5 in dllname () from /usr/lib/libwsock32.so #1 0x08121821 in StreamServerPort () #2 0x08150996 in PostmasterMain () #3 0x08123353 in main () #4 0x4010e7ee in __libc_start_main () from /lib/libc.so.6 elein@varlena.com
That line is certainly strange: #0 0x40099ac5 in dllname () from /usr/lib/libwsock32.so When you run configure, it says you are on Linux, right? My guess is that gdb is getting confused because there is no dllopen call in StreamServerPort(). --------------------------------------------------------------------------- elein wrote: > > SuSE 7.3, PostgreSQL cvshead (7.4) > > This is as far as I've gotten with 7.4. > It is built and configured like my 7.3 installation > on the same machine. I have built from CVS previously. > And the production sources always builds very nice and clean. > > Nothing runs. gbd output is below. It is balking > on loading a suspect library. > > I ordinarily would suspect some minor permission > or configuration issue, but I cannot see one. > > Joe Conway suggested it might be a problem with the > the windows configuration changes. > > My machine has WINE on it (windows) although > I never use it. There really should be no > reason for anything to be trying to link > in a windows socket library. > > I don't have time to follow it further right now. > But if someone else has a similar configuration > and/or wine installed, it would be very helpful > to compare 7.4 installations. I can provide > config arguments and log if someone would find > them useful. > > >gdb bin/postmaster > GNU gdb 20010316 > Copyright 2001 Free Software Foundation, Inc. > GDB is free software, covered by the GNU General Public License, and you are > welcome to change it and/or distribute copies of it under certain conditions. > Type "show copying" to see the conditions. > There is absolutely no warranty for GDB. Type "show warranty" for details. > This GDB was configured as "i386-suse-linux"... > (gdb) run -i -D /local/pghead/data > Starting program: /local/pghead/bin/postmaster -i -D /local/pghead/data > > Program received signal SIGSEGV, Segmentation fault. > 0x40099ac5 in dllname () from /usr/lib/libwsock32.so > (gdb) bt > #0 0x40099ac5 in dllname () from /usr/lib/libwsock32.so > #1 0x08121821 in StreamServerPort () > #2 0x08150996 in PostmasterMain () > #3 0x08123353 in main () > #4 0x4010e7ee in __libc_start_main () from /lib/libc.so.6 > > > elein@varlena.com > > ---------------------------(end of broadcast)--------------------------- > TIP 7: don't forget to increase your free space map settings > -- Bruce Momjian | http://candle.pha.pa.us pgman@candle.pha.pa.us | (610) 359-1001+ If your life is a hard drive, | 13 Roberts Road + Christ can be your backup. | Newtown Square, Pennsylvania19073
elein <elein@norcov.com> writes:
> This is as far as I've gotten with 7.4.
Would you rebuild with --enable-debug (perhaps also --enable-cassert)
so that the gdb backtrace is more informative?
Also, it seems likely that the issue is in or around the recently-added
IPv6 support, so I'd suggest using CVS tip or last night's snapshot
rather than the beta1 tarball. We've already made some portability
fixes there since beta1.
regards, tom lane
cassert was on. Now debug is on, too. I updated from cvs-head just now. configure knows it is a linux box. Should it be trying to link to libwsock32.so or not? If this is a legitimate link, then the problem is different than if it is trying to link it in erroneously. --elein The is the top of the config.log --------------------- This file contains any messages produced by compilers while running configure, to aid debugging if configure makes a mistake. It was created by PostgreSQL configure 7.4beta1, which was generated by GNU Autoconf 2.53. Invocation command line was $ ./configure --prefix=/local/pghead --with-perl --with-python --enable-depend --enable-cassert --enable-d ebug ## --------- ## ## Platform. ## ## --------- ## hostname = cookie uname -m = i686 uname -r = 2.4.16-4GB uname -s = Linux uname -v = #1 Mon Apr 15 08:57:26 GMT 2002 --------------------- $ gdb postmaster GNU gdb 20010316 Copyright 2001 Free Software Foundation, Inc. GDB is free software, covered by the GNU General Public License, and you are welcome to change it and/or distribute copies of it under certain conditions. Type "show copying" to see the conditions. There is absolutely no warranty for GDB. Type "show warranty" for details. This GDB was configured as "i386-suse-linux"... (gdb) run -i -D /local/pghead/data Starting program: /local/pghead/bin/postmaster -i -D /local/pghead/data Program received signal SIGSEGV, Segmentation fault. 0x40099ac5 in dllname () from /usr/lib/libwsock32.so (gdb) bt #0 0x40099ac5 in dllname () from /usr/lib/libwsock32.so #1 0x081218b1 in StreamServerPort (family=0, hostName=0x0, portNumber=5432, unixSocketName=0x82d6e68 "", ListenSocket=0x829b420,MaxListen=10) at pqcomm.c:279 #2 0x08150a26 in PostmasterMain (argc=4, argv=0x82cfae8) at postmaster.c:765 #3 0x081233e3 in main (argc=4, argv=0xbffff414) at main.c:215 #4 0x4010e7ee in __libc_start_main () from /lib/libc.so.6 On Mon, Aug 11, 2003 at 11:54:39AM -0400, Tom Lane wrote: > elein <elein@norcov.com> writes: > > This is as far as I've gotten with 7.4. > > Would you rebuild with --enable-debug (perhaps also --enable-cassert) > so that the gdb backtrace is more informative? > > Also, it seems likely that the issue is in or around the recently-added > IPv6 support, so I'd suggest using CVS tip or last night's snapshot > rather than the beta1 tarball. We've already made some portability > fixes there since beta1. > > regards, tom lane >
elein <elein@norcov.com> writes:
> configure knows it is a linux box.
> Should it be trying to link to libwsock32.so
> or not? If this is a legitimate link, then
> the problem is different than if it is trying
> to link it in erroneously.
configure is unconditionally including libwsock32 if it can find one.
AFAICT from the CVS logs, this was only expected to happen on win32
(Bruce, that was your commit, configure.in v1.250; please confirm).
So it would probably make sense to not look for libwsock32 unless
PORTNAME is "win32".
I take it you actually have a libwsock32? What's it supposed to do?
regards, tom lane
Yes, I actually have a libwsock32 because my system has wine on it. Wine is a windows emulator. So the assumption that any system with that file is a windows system will break on systems with windows emulators. It sounds like Joe's guess on this was right. --elein On Mon, Aug 11, 2003 at 01:29:19PM -0400, Tom Lane wrote: > elein <elein@norcov.com> writes: > > configure knows it is a linux box. > > Should it be trying to link to libwsock32.so > > or not? If this is a legitimate link, then > > the problem is different than if it is trying > > to link it in erroneously. > > configure is unconditionally including libwsock32 if it can find one. > AFAICT from the CVS logs, this was only expected to happen on win32 > (Bruce, that was your commit, configure.in v1.250; please confirm). > So it would probably make sense to not look for libwsock32 unless > PORTNAME is "win32". > > I take it you actually have a libwsock32? What's it supposed to do? > > regards, tom lane >
elein <elein@norcov.com> writes:
> Yes, I actually have a libwsock32 because my
> system has wine on it. Wine is a windows
> emulator.
And they drop windows-only libraries into /usr/lib? Yech.
Anyway, I can't see a need to include libwsock32 on non-win32 platforms.
Will modify configure.
regards, tom lane
elein <elein@norcov.com> writes:
> It sounds like Joe's guess on this was right.
I've committed this fix in configure.in:
***************
*** 631,637 **** AC_CHECK_LIB(gen, main) AC_CHECK_LIB(PW, main) AC_CHECK_LIB(resolv, main)
- AC_CHECK_LIB(wsock32, main) AC_SEARCH_LIBS(getopt_long, [getopt gnugetopt]) # QNX: AC_CHECK_LIB(unix, main)
--- 636,641 ----
***************
*** 645,650 ****
--- 649,659 ---- AC_SEARCH_LIBS(fdatasync, [rt posix4]) # Cygwin: AC_CHECK_LIB(cygipc, shmget)
+ # WIN32:
+ if test "$PORTNAME" = "win32"
+ then
+ AC_CHECK_LIB(wsock32, main)
+ fi if test "$with_readline" = yes; then PGAC_CHECK_READLINE
regards, tom lane
On Mon, 2003-08-11 at 13:44, elein wrote: > Yes, I actually have a libwsock32 because my > system has wine on it. Wine is a windows > emulator. > Wine Is Not an Emulator :-) Robert Treat -- PostgreSQL :: The Enterprise Open Source Database
Yes, this is the right fix. I never suspected wsock32 would exist on a non-MS WIn machine. --------------------------------------------------------------------------- Tom Lane wrote: > elein <elein@norcov.com> writes: > > It sounds like Joe's guess on this was right. > > I've committed this fix in configure.in: > > *************** > *** 631,637 **** > AC_CHECK_LIB(gen, main) > AC_CHECK_LIB(PW, main) > AC_CHECK_LIB(resolv, main) > - AC_CHECK_LIB(wsock32, main) > AC_SEARCH_LIBS(getopt_long, [getopt gnugetopt]) > # QNX: > AC_CHECK_LIB(unix, main) > --- 636,641 ---- > *************** > *** 645,650 **** > --- 649,659 ---- > AC_SEARCH_LIBS(fdatasync, [rt posix4]) > # Cygwin: > AC_CHECK_LIB(cygipc, shmget) > + # WIN32: > + if test "$PORTNAME" = "win32" > + then > + AC_CHECK_LIB(wsock32, main) > + fi > > if test "$with_readline" = yes; then > PGAC_CHECK_READLINE > > > regards, tom lane > -- Bruce Momjian | http://candle.pha.pa.us pgman@candle.pha.pa.us | (610) 359-1001+ If your life is a hard drive, | 13 Roberts Road + Christ can be your backup. | Newtown Square, Pennsylvania19073
I blame SuSE. Thank you for the fix and confirmation of the problem. elein On Mon, Aug 11, 2003 at 01:53:31PM -0400, Tom Lane wrote: > elein <elein@norcov.com> writes: > > Yes, I actually have a libwsock32 because my > > system has wine on it. Wine is a windows > > emulator. > > And they drop windows-only libraries into /usr/lib? Yech. > > Anyway, I can't see a need to include libwsock32 on non-win32 platforms. > Will modify configure. > > regards, tom lane >