Suppress error output to screen or fix a problem?

[PabloTwo]

Freshly off the 'sed' issue on my rpm packaging project (I found another instance I needed sed to alter a file and was able to do so with the earlier help, thank you). I'm down to what would seem like one last tweak of the spec file. The rpm package builds and installs just fine, but during the install, there is a repeated error message, which doesn't seem to adversely the install, so I'm wondering if I either need to suppress this output to screen or if I need to correct a problem someplace to prevent the error messages in the first place. The install:

 BASH:~/-> sudo rpm --nosignature -ih rmsgw-2.0.2-111.i386.rpm 
########################################### [100%]
/usr/sbin/nscd: relocation error: /usr/sbin/nscd: symbol __nss_services_lookup, version GLIBC_PRIVATE not defined in file libc.so.6 with link time reference
/usr/sbin/nscd: relocation error: /usr/sbin/nscd: symbol __nss_services_lookup, version GLIBC_PRIVATE not defined in file libc.so.6 with link time reference
/usr/sbin/nscd: relocation error: /usr/sbin/nscd: symbol __nss_services_lookup, version GLIBC_PRIVATE not defined in file libc.so.6 with link time reference
/usr/sbin/nscd: relocation error: /usr/sbin/nscd: symbol __nss_services_lookup, version GLIBC_PRIVATE not defined in file libc.so.6 with link time reference
########################################### [100%]

And the line of code that I believe to be at the root of the issue (in the %pre section of the spec file), but not absolutely sure:

BASH:~/-> cat files/specfiles/rmsgw-2.0.2-111.spec | grep -A2 "%pre"

---<snip>---

%pre
getent passwd rmsgw >/dev/null || /usr/sbin/useradd -r rmsgw >/dev/null || :

Do I need more than the >/dev/null bits to get rid of this ouput?

I anticipate a lot of questions being raised here... yeah, the code author "insists" all files installed in this package have rmsgw:rmsgw (user:group) ownership, so creating the user 'rmsgw' is necessary to fullfill that requirement. I chose the '-r' option to the useradd command to negate having to deal with the creation of /home/rmsgw and /var/spool/mail/rmsgw, which aren't needed. rpmlint isn't happy with this arrangement either.

Any ideas, suggestions?

Paul

Edit: Looking at the man page for nscd causes me to suspect perhaps this is occurring due to the fact that the nscd service is not running on my machine?, which it isn't.

Edit#2: Doing 'sudo service nscd start' gave me the exact same error message. Tried from the GUI Services app and same thing. Uninstalled and reinstalled nscd and rebooted. Same error. Looks like something amiss in my glibc-2.8-3.i686 perhaps.

Edit#3:

BASH:~/-> rpm -V glibc
..?.....    /usr/sbin/glibc_post_upgrade.i686

Anyone know how to decipher this? I'm running F8. My glibc package is an update from the F9 repos. Looks like there is no problem with the spec file, only the OP's hybridized Fedora Ah.. got it... ? in 3rd position indicates: MD5 sum differs. Looks like I have to force a reinstall of glibc.

Edit#4: Once I replaced the glibc-2.8-3.i686 with glibc-2.8-3.i386, to match the rest of the installed glibc group, and updated nscd-7.x-x to nscd-2.8-3.i386, which is part of that 'group', everything is roses again.

[regulatre]

Those library errors from nscd are a little suspicious.

If you just want to supress them, you'll need to redirect stdERR in addition to stdOUT (which you're already getting with >/dev/null).

mycommand > /dev/null

mycommand > /dev/null 2>&1

The second command above takes STDERR and STDOUT and pipes them together into /dev/null so they don't go to the screen. Basically the 2 represents stderr, and the 1 is stdout. >/dev/null is a shortcut for 1>/dev/null.

It might be good for those errors to be sent to the screen though... Especially a library error. I'd look into that.

[PabloTwo]

Thank you for the explanation about the "2>&1" bit. I'd been looking for just that recently. And you are right that it's better to see the errors in order to fix the problem. As stated in my "Edit#4" in the previous post above, I removed the errors generated by fixing the problem that was causing them. It took using the 'getent' command to reveal the problem in the first place.