This is the mail archive of the
cygwin@cygwin.com
mailing list for the Cygwin project.
RE: postgresql question
- From: "Ralf Habacker" <Ralf dot Habacker at freenet dot de>
- To: "Jason Tishler" <jason at tishler dot net>
- Cc: "Cygwin" <cygwin at cygwin dot com>
- Date: Sat, 9 Nov 2002 03:21:45 +0100
- Subject: RE: postgresql question
> On Fri, Nov 08, 2002 at 09:50:45AM +0100, Ralf Habacker wrote:
> > > -----Original Message-----
> > > From: Jason Tishler [mailto:jason@tishler.net]
> > > Sent: Thursday, November 07, 2002 7:48 PM
> > >
> > But some pgcyrpto functions returns different crypted hashes as under
> > linux for example and that is the reason why the below mentioned
> > regression test fails.
>
> Could this be a binary versus text issue? See below...
>
> > > Does proftpd work for you under Cygwin? When I tried, I discovered
> > > many issues even though it built OOTB.
> >
> > What do you mean especially ?
>
> So far I have found the following issues with Cygwin ProFTPD 1.2.6:
I'm used the recent cvs release
>
> o inaccuracies in the README.cygwin
> o build problems due to missing EXEEXTs in Makefile.in
no problem
> o password authentication problems (need to use cygwin_logon_user())
> o ls fails
no problem
> o get reports inaccurate file sizes
> o numerous setuid(0) calls
there is only one place where i have to comment it out
Index: main.c
===================================================================
RCS file: /cvsroot/proftp/proftpd/src/main.c,v
retrieving revision 1.128
diff -u -3 -p -B -r1.128 main.c
--- main.c 28 Oct 2002 16:51:50 -0000 1.128
+++ main.c 8 Nov 2002 23:42:04 -0000
@@ -3097,6 +3097,7 @@ int main(int argc, char *argv[], char **
* a portable fashion *gah!*
*/
+#ifndef __CYGWIN__
if (geteuid() != daemon_uid) {
log_pri(PR_LOG_ERR, "unable to set uid to %lu, current uid: %lu",
(unsigned long)daemon_uid,(unsigned long)geteuid());
@@ -3108,6 +3109,7 @@ int main(int argc, char *argv[], char **
(unsigned long)daemon_gid,(unsigned long)getegid());
exit(1);
}
+#endif
/* Install a signal handlers/abort handler */
install_signal_handlers();
>
> > [snip]
> > > What is the above? It looks like a PostgreSQL regression test
> > > output, but I have never seen the above when I run the regression
> > > test.
> >
> > $ cd <postgresql-source-root>/contrib/pgcrypto
> > make installcheck
>
> Doh! I tried make test yesterday -- I should know better!
>
> BTW, all tests succeed for me:
>
> test init ... ok
> test md5 ... ok
> test sha1 ... ok
> test hmac-md5 ... ok
> test hmac-sha1 ... ok
> test blowfish ... ok
> test rijndael ... ok
> test crypt-des ... ok
> test crypt-md5 ... ok
> test crypt-blowfish ... ok
> test crypt-xdes ... ok
>
> > As i said above, it seems to me, that there are different md5
> > implementations at all. It's very confusing. In short, postgresql
> > create different md5 keys under cygwin and linux
>
> See above.
Im using binary mounts
$ mount
c:\programme\cygwin\bin on /usr/bin type system (binmode)
c:\programme\cygwin\lib on /usr/lib type system (binmode)
c:\programme\cygwin on / type system (binmode)
c: on /c type user (binmode,noumount)
d: on /d type user (binmode,noumount)
>
> > > BTW, I'm still looking for a cygipc beta tester:
> > >
> > > http://cygwin.com/ml/cygwin/2002-11/msg00136.html
> > >
> > > Are you willing to help me out?
> > >
> > [snip]
> >
> > Have I only update the ipc-daemon or have I to rebuild all the client
> > applications ?
> >
> > The patches let me assume that this affect only the daemon.
>
> No, the patch actually affects the library, libcygipc.a. Unfortunately,
> you will have to build all (or at least some) of the client
> applications.
Jason, I have inspect the kde (kdelibs/kdebase) sources about this issue and
found, that any code which uses shmget (semget is never used) handles return
code correctly (-1 = error code) so that adding the ENOSYS return code does not
affect kde.
Next I have taken the qt sources and found the same results.
Next I have taken a look into the xfree sources and found that the Xserver
handles the return codes valid.
Nice code below:
xc/programs/Xserver/hw/xfree86/os-support/linux/int10/linux.c
</snip>
if ((high_mem = shmget(counter++, HIGH_MEM_SIZE,
IPC_CREAT | SHM_R | SHM_W)) == -1) {
if (errno == ENOSYS)
xf86DrvMsg(screen, X_ERROR, "shmget error\n Please reconfigure"
" your kernel to include System V IPC support\n");
goto error1;
<snip>
Because the linux implementation of shmget()/semget() returns already ENOSYS in
case of "no shm implementation" errors and because kde code runs already very
long on linux, i believe that there is no problem with this patch.
So from my view there are no reason, why this patch does be go into the offical
cygipc release.
Jason, I haven't recompiled kde with this release, because this job isn't
possible for me at this moment. A whole recompile of kde and testing is a job of
several days and I will do that, when the next kde release is going to be
released.
BTW: I haven't recongized you patch before. So please sorry if this have delayed
your release planing of the postgresql port.
You have done a great job.
Regards
Ralf
--
Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple
Bug reporting: http://cygwin.com/bugs.html
Documentation: http://cygwin.com/docs.html
FAQ: http://cygwin.com/faq/