Обсуждение: IpcMemoryDetach fails under Cygwin on Win98
I'm frantically trying to get PostgreSQL 7.0.3 installed on my laptop so I can work on my database while on holiday next week. I've got it to compile fine with the latest version of Cygwin, with cygipc-1.08, however, when I start postmaster, it fails: $ ipc-daemon & $ postmaster -i & [2] 528923 $ G:\CYGW\USR\LOCAL\PGSQL\BIN\POSTGRES.EXE: *** recreate_mmaps_after_for k_failed Startup failed - abort NOTICE: IpcMemoryDetach: shmdt(0x0x82f6a000): Invalid argument NOTICE: IpcMemoryDetach: shmdt(0x0x86ba1000): Invalid argument NOTICE: IpcMemoryDetach: shmdt(0x0x82c50000): Invalid argument I have applied the patch indicated in a previous thread for cygipc, but the error still persists. I note from the mailing list that someone has had PostgreSQL working under Win98 before - I would be extremely grateful if someone could help me get this working. I can't seem to get gdb to work properly under Cygwin (would not find the source files), so I was unable to step through the code and see what the problem is. -- Jamie Walker "My dear boy, forget about the motivation. jamie@sagaxis.co.uk Just say the lines and don't trip over the http://www.sagaxis.co.uk/ furniture." (Noel Coward)
On Tue, Feb 06, 2001 at 03:04:10PM +0100, Christoph Appel wrote:
> I built 7.0.3 using the KevinLo-Howto, as well as 7.1b4 out-of-the-box.
> My environment is PIII 600 Mhz with 64MB RAM, Win98SE, cygwin-1.1.7,
> cygipc-1.08, gcc-2.95-5.
>
> The regression tests however yield some problems: I was able to pass
> (nearly, under 7.0.3 horology failed) all tests, but not all at a time.
> When running make runtest, the test always hangs somewhere. To then be
> able to continue testing, I have to shutdown the server and even kill
> ipc-daemon and restart both. By running only fewer tests at a time, I
> nonetheless was able to eventually pass all (but horology under 7.0.3)
> tests.
>
> The Perl package DBD-Pg-0.95 builds and passes all tests with both 7.0.3
> and 7.1b4.
Please see item #2 in the following:
http://postgresql.readysetnet.com/mhonarc/pgsql-ports/2001-01/msg00066.html
Does the description match what you are observing?
Please respond to the list too.
Thanks,
Jason
--
Jason Tishler
Director, Software Engineering Phone: +1 (732) 264-8770 x235
Dot Hill Systems Corp. Fax: +1 (732) 264-8798
82 Bethany Road, Suite 7 Email: Jason.Tishler@dothill.com
Hazlet, NJ 07730 USA WWW: http://www.dothill.com
Hi there, >> Can other Win98/Cygin people comment on when it >> started working on Win98? > Unfortunately, I never ran the regression tests on Windows 9X or Me. I > also will not have access to a 9X or Me machine for a while. Can > someone run the regression tests on such a machine and report back to the > list? > Cygwin runs pretty much the same on 9X/Me as on NT/2000 except for a few > rough edges and some perceived pokiness. I think that the services that > PostgreSQL needs from Cygwin should be okay on 9X/Me but the regression > test results will give us the definitive answer. I built 7.0.3 using the KevinLo-Howto, as well as 7.1b4 out-of-the-box. My environment is PIII 600 Mhz with 64MB RAM, Win98SE, cygwin-1.1.7, cygipc-1.08, gcc-2.95-5. The regression tests however yield some problems: I was able to pass (nearly, under 7.0.3 horology failed) all tests, but not all at a time. When running make runtest, the test always hangs somewhere. To then be able to continue testing, I have to shutdown the server and even kill ipc-daemon and restart both. By running only fewer tests at a time, I nonetheless was able to eventually pass all (but horology under 7.0.3) tests. I already noticed http://postgresql.readysetnet.com/mhonarc/pgsql-ports/2001-01/msg00066.html but this doesn't match my observation. The test as well as the server just hang without producing any further error messages. The Perl package DBD-Pg-0.95 builds and passes all tests with both 7.0.3 and 7.1b4. Bye, Chris
Chris, On Tue, Feb 06, 2001 at 03:44:57PM +0100, Christoph Appel wrote: > The test as well as the server just hang without producing any further > error messages. On Mon, Jan 15, 2001 at 11:37:55PM -0500, Jason Tishler wrote: > The tests usually hang during the "parallel group (18 tests)" test > right after numerology. By "hang," I mean that the original postmaster > is still running, but there are no postmaster children, and there are > some number of psql processes hanging around. Please do a ps. Are you sure that you don't see some psql processes hanging around? Use NT's TaskManager (specifically the Performance tab), what is the memory usage during the regression tests. Are you running out of memory? Thanks, Jason -- Jason Tishler Director, Software Engineering Phone: +1 (732) 264-8770 x235 Dot Hill Systems Corp. Fax: +1 (732) 264-8798 82 Bethany Road, Suite 7 Email: Jason.Tishler@dothill.com Hazlet, NJ 07730 USA WWW: http://www.dothill.com
Tried 7.1beta4, and the same thing happens... Added more memory, and it takes longer to hang. Maybe with 256mb it can finish. BTW, this problem is from the regression test, or some application using the DB Engine could hang also? Jason Tishler <Jason.Tishler@dothill.com> escreveu nas not�cias de mensagem:20010206092051.H1084@dothill.com... > On Tue, Feb 06, 2001 at 03:04:10PM +0100, Christoph Appel wrote: > > I built 7.0.3 using the KevinLo-Howto, as well as 7.1b4 out-of-the-box. > > My environment is PIII 600 Mhz with 64MB RAM, Win98SE, cygwin-1.1.7, > > cygipc-1.08, gcc-2.95-5. > > > > The regression tests however yield some problems: I was able to pass > > (nearly, under 7.0.3 horology failed) all tests, but not all at a time. > > When running make runtest, the test always hangs somewhere. To then be > > able to continue testing, I have to shutdown the server and even kill > > ipc-daemon and restart both. By running only fewer tests at a time, I > > nonetheless was able to eventually pass all (but horology under 7.0.3) > > tests. > > > > The Perl package DBD-Pg-0.95 builds and passes all tests with both 7.0.3 > > and 7.1b4. > > Please see item #2 in the following: > > http://postgresql.readysetnet.com/mhonarc/pgsql-ports/2001-01/msg00066.html > > Does the description match what you are observing? > > Please respond to the list too. > > Thanks, > Jason > > -- > Jason Tishler > Director, Software Engineering Phone: +1 (732) 264-8770 x235 > Dot Hill Systems Corp. Fax: +1 (732) 264-8798 > 82 Bethany Road, Suite 7 Email: Jason.Tishler@dothill.com > Hazlet, NJ 07730 USA WWW: http://www.dothill.com
Hi Jason, > Chris, > > On Tue, Feb 06, 2001 at 03:44:57PM +0100, Christoph Appel wrote: > > The test as well as the server just hang without producing any further > > error messages. > > On Mon, Jan 15, 2001 at 11:37:55PM -0500, Jason Tishler wrote: > > The tests usually hang during the "parallel group (18 tests)" test > > right after numerology. By "hang," I mean that the original postmaster > > is still running, but there are no postmaster children, and there are > > some number of psql processes hanging around. > > Please do a ps. Are you sure that you don't see some psql processes > hanging around? Well, the one 'hanging' psql process is there (I'm talking of the serialized test, so this looks reasonable to me). I then terminate the test via CTRL-C. The postmaster is also still there, but without any children. The postmaster can be shutdown via 'pg_ctl stop' (without producing any further error messages; I got rid of the IPC err messages by applying the sem and shm patches I noticed in the mailing list). Like the ipc-daemon, the postmaster has to be shut down and restarted in order to be able to run the test again. Without restarting the two, a new run of the test will hang right at the drop old regression database. In this case there is a postgres child hanging around which does not even react on a TERM, and postmaster will not shut down before KILLing the child. (With postmaster I mean the original postgres process.) Under 7.0.3 the test hangs at 'create_aggregate'. When taking out some .sql file before 'create_aggregate' it will pass 'create_aggregate' but hang somewhere later. By taking out a number of .sql files I will eventually pass all (remaining) tests. > Use NT's TaskManager (specifically the Performance tab), what is the memory > usage during the regression tests. Are you running out of memory? Sorry, no NT TaskManager under Win98. Do you know some tool or method I can check this? > Thanks, > Jason Hth, Chris
Chris, On Wed, Feb 07, 2001 at 10:02:44AM +0100, Christoph Appel wrote: > Well, the one 'hanging' psql process is there (I'm talking of the > serialized test, so this looks reasonable to me). I then terminate the > test via CTRL-C. Please try a "make installcheck" -- this will run the tests sequentially instead of in parallel. I'm doubtful that this will work since others have already tried it and were unsuccessful, but since it is easy it is worth a shot. Remember to install PostgreSQL and start postmaster (and ipc-daemon) before you run this version of the regression test. > postmaster can be shutdown via 'pg_ctl stop' (without producing any > further error messages; I got rid of the IPC err messages by applying > the sem and shm patches I noticed in the mailing list). Note that these patches are no longer necessary when you use the latest Cygwin CVS or Cygwin 1.3.0 (which is yet to be released). > > Use NT's TaskManager (specifically the Performance tab), what is the memory > > usage during the regression tests. Are you running out of memory? > > Sorry, no NT TaskManager under Win98. Doh! > Do you know some tool or method I can check this? I quickly searched the MS MSDN/KB but came up empty. As the Core Cygwin developers have said many times, "I hate Windows 9X/Me!" Jason -- Jason Tishler Director, Software Engineering Phone: +1 (732) 264-8770 x235 Dot Hill Systems Corp. Fax: +1 (732) 264-8798 82 Bethany Road, Suite 7 Email: Jason.Tishler@dothill.com Hazlet, NJ 07730 USA WWW: http://www.dothill.com