homespun

MS Word Find and Replace tricks

references:

1. http://www.csgnetwork.com/asciiset.html
2. http://www.techrepublic.com/forum/questions/101-213548/paragraph-code-vs-13-code-in-ms-word
3. http://office.microsoft.com/en-us/support/putting-regular-expressions-to-work-in-word-HA001087304.aspx

When I use Find and Replace in MS Word (I'm using Word 2010, but this will work in 2007 and 2003), I use ^p and ^t to find paragraph and tab marks, respectively. But, if I'm doing wildcard searches, then those 2 codes don't work.

I was reading up on regular expressions in MS Word's F&R (ref. 3) and they mentioned that you could use ^13 in regular expressions instead of ^p. There were some warnings in there as well. The paragraph mark in Word contains the formatting information for the preceding paragraph, whereas the ^13 is merely a carriage return.

However, "13" is the ANSI decimal code for a carriage return (CR), so I looked up some other ANSI control codes (ref. 1). The ANSI decimal code for the horizontal tab character is "9". Sure enough, in Word's F&R, I can substitute ^9 for ^t when using regular expressions.

When used carefully, this allows me to do (almost) all of my F&R using wildcard expressions. This solves a long-term gripe of mine. Great!

udev network persistence and VMware issues

So, modern linuxes use udev to manage devices. One of the advantages of udev is that it "supports persistent device naming" and yes, in general this is a good thing. However, it comes back to bite you when dealing with the virtual world. For example...

I have VMware virtual machine with Centos 6.2 installed. The default NIC is named eth0, nothing surprising here. I have installed VMware tools, applied all of the Centos yum updates and now I clone my VM rather than repeating the install process.

Networking won't start on the new clone. eth0 is not present or can't be found. The answer is pretty simple, but it still took me awhile to figure it out and get my head around the nuances. The root cause of the problem is that the clone has a different HWADDR, or MAC address, than the original.VM. That makes total sense. You wouldn't want two devices on the same network to have the same MAC address - now there's a recipe for disaster.

But, udev has made a persistent connection to the original vNIC and stored the original MAC address in /etc/udev/rules.d/70-persistent-net.rules. At boot, udev can't find the original eth0 and reports "Device eth0 does not seem to be present, delaying initialization". My Centos 6.2 does discover a new vNIC but it's not configured or activated:

# ls /sys/class/net
eth1 lo

There are probably lots of ways to fix the situation. This one is just brute force:

1. as root, delete the persistent rule file
   • rm -f /etc/udev/rules.d/*-net.rules
2. run ifconfig, note the HWaddr of eth1.
3. edit /etc/sysconfig/networking-scripts/ifcfg-eth0 and replace the HWADDR with the correct MAC address from step 2.
4. reboot. A new 70-persistent-net.rules file will be created automatically.

This message may only work for RedHat, Fedora, or Centos. I haven't tried it with other distros. Ubuntu also uses udev, so it may work.

EDIT: by the way, on Centos 6, there are 3 hard links to /etc/sysconfig/network-scripts/ifcfg-eth0:

[root@x-serv ~]# find / -xdev -samefile /etc/sysconfig/network-scripts/ifcfg-eth0
/etc/sysconfig/network-scripts/ifcfg-eth0
/etc/sysconfig/networking/profiles/default/ifcfg-eth0
/etc/sysconfig/networking/devices/ifcfg-eth0

My method preserves all of the hard links. If you delete any one of those files (I used to delete /etc/sysconfig/network-scripts/ifgcfg-eth0, for example), then the other 2 will still remain and could cause problems down the road.

libncurses.so.5 undefined symbol _nc_set_no_padding

I was looking to diagnose a high load problem today. For this, I typically use atop. However, I got this output instead...

[root@myserver ~] # atop: symbol lookup error: /lib/libncurses.so.5: undefined symbol: _nc_set_no_padding

I'm running CentOS 6.2, with atop installed from the EPEL repository so I figured it should work. I started looking for another answer. Google is your friend in all such cases, and I found this bugzilla report which implicates zend. Since we are in fact using zend, I thought I would verify the issue.

[root@x-web-01v lib]# which atop
/usr/bin/atop
[root@x-web-01v lib]# readelf -d $(which atop) | fgrep NEEDED
0x00000001 (NEEDED)                     Shared library: [libncurses.so.5]
0x00000001 (NEEDED)                     Shared library: [libm.so.6]
0x00000001 (NEEDED)                     Shared library: [libz.so.1]
0x00000001 (NEEDED)                     Shared library: [libc.so.6]
0x00000001 (NEEDED)                     Shared library: [libtinfo.so.5]
[root@x-web-01v lib]# ldd $(which atop)
        linux-gate.so.1 =>  (0x0016b000)
        libncurses.so.5 => /lib/libncurses.so.5 (0x00b7e000)
        libm.so.6 => /lib/libm.so.6 (0x00e5d000)
        libz.so.1 => /lib/libz.so.1 (0x0072e000)
        libc.so.6 => /lib/libc.so.6 (0x0016c000)
       libtinfo.so.5 => /usr/local/zend/lib/libtinfo.so.5 (0x00110000)
        libdl.so.2 => /lib/libdl.so.2 (0x00e21000)
        /lib/ld-linux.so.2 (0x004f6000)
[root@x-web-01v lib]# nm -D /lib/libncurses.so.5 | fgrep _nc_set_no_padding
         U _nc_set_no_padding

So, sure enough, zend has hijacked libtinfo.so.5. How to fix this - edit /etc/ld.so.conf and add the path to /lib (or /lib64 for 64-bit OS) at the top:

[root@myserver ~]# cat /etc/ld.so.conf
/lib/
/usr/lib
include ld.so.conf.d/*.conf

I added /usr/lib for good measure. Run ldconfig to make the changes live.

analog-6.0.4 on Centos6.2

I'm building up a new server on Centos 6.2 to replace an older RedHat 5.6 (Tikanga) box. And I needed to install analog logfile analyzer as part of the build. My 1st hurdle was finding the Analog website :) The site that I'm used to - www.analog.cx - doesn't seem to exist any longer. However, the sourceforge site is working just fine.

On the download page, there are 2 links to pre-compiled RedHat Linux binaries. I went to site 1 and downloaded the 6.0.4 SRPM, since I needed the i686 version for my build.

I tried to recompile the rpm from source:

# rpmbuild --rebuild analog-6.0.4-1.src.rpm
Installing analog-6.0.4-1.src.rpm
error: File /root/rpmbuild/SOURCES/analog-%{PACKAGE_VERSION}.tar.bz2: No such file or directory

I used rpmlint to look for errors in the spec file that was created:

# rpmlint rpmbuild/SPECS/analog.spec
rpmbuild/SPECS/analog.spec:8 W: non-standard-group Utilities
rpmbuild/SPECS/analog.spec:10 W: hardcoded-path-in-buildroot-tag /tmp/anlg-root
rpmbuild/SPECS/analog.spec:17 W: setup-not-quiet
rpmbuild/SPECS/analog.spec:66 W: macro-in-%changelog %defattr
rpmbuild/SPECS/analog.spec:74 W: macro-in-%changelog %config
rpmbuild/SPECS/analog.spec: W: invalid-url Source0: http://www.analog.cx/analog-%{PACKAGE_VERSION}.tar.gz2 <urlopen error [Errorno -2] Name or service not known>
0 packages and 1 specfiles checked; 0 errors, 6 warnings.

I edited the analog.spec file and made the following changes.

• replaced all instances of 'PACKAGE_VERSION' with 'version'
• line 8 - changed group to Applications/Internet (this works on CentOS 6.2 - ymmv)
• line 10 - change BuildRoot value to %{_tmppath}/%{name}-%{version}
• line 17 - added "-q" flag to setup
• line 64 - changed "%changelog" to "%%changelog" to resolve the 2 macro warnings
• changed the URL

Then, I was able to build and, more importantly, install the rpm.

# rpmbuild -ba rpmbuild/SPECS/analog.spec
# rpm -ihv rpmbuild/RPMS/i686/analog-6.0.4-1.i686.rpm

Mounting CIFS shares in fstab - RHEL5

There's a number of references to mounting cifs shares in /etc/fstab out there. I thought the swerdna article was very helpful. Although the author talks about OpenSuSE, it was close enough for most of my RHEL5 needs.

I ran into the problem of a late-starting network. I tried the techniques in the Troubleshooting section, but none of them seemed to work. Although man mount talked about the _netdev option, it didn't work on my system - RHEL 5.7 - and gave me an error during boot.

I also tried the trick about using a crontab entry - also failed to work. The last workaround was to use after.local to run mount -a. However, RedHat doesn't use after.local. The answer was to put the command in /etc/rc.d/rc.local. This file does the same thing for RedHat that after.local does for OpenSuSE.

Lastly, I needed lighttpd to start using a document root on the CIFS share. In short, the daemon wouldn't start if the mount wasn't present. So, I added service lighttpd restart after the mount command in /etc/rc.d/rc.local. Worked like a charm!

Bonding NICs in RHEL 5.5

I ran into something yesterday. I installed RHEL 5.5, then set about bonding two nics together. When the bond was set to a static IP in a different LAN segment, the server wouldn't finish booting. It was somewhat responsive to keyboard input (Enter caused the screen to advance one line) but that was about it.

I thought the install was bad, so I re-ran the installation. As long as the nics weren't bonded, the server booted fine. I installed updates from the RH repos, etc etc. But as soon as I bonded the nics again, it wouldn't reboot.

I racked the server, plugged the nics into the proper subnet - and the thing booted fine.