| Sysadmin Log |
[Feb. 15th, 2009|12:00 pm] |
Copied ssh keys over from sleipnir.
Installed vim-full, vim-doc, manpages-dev, manpages-posix, manpages-posix-dev.
Installed fakeroot build-essential crash kexec-tools makedumpfile git-core libncurses5 libncurses5-dev per [1].
apt-get build-dep linux per [1].
apt-get build-dep r-base
apt-get install synaptic pan abakus git-gui git-doc abakus kcalc firefox ubufox mozilla-plugin-vlc xpdf-reader tetex-bin xfig r-base |
|
|
| Kubuntu Jaunty: Initial impressions |
[Feb. 15th, 2009|10:45 am] |
Plasmoids
1) No way to close plasmoids (Calendar) by clicking on them. You have to figure out what caused the plasmoid to come into existence in the first place and click on that -- in this case, the digital clock on the taskbar.
2) No way to get rid of persistent plasmoids by clicking on them. You have to figure out that "add" means "add or remove".
3) Cannot resize plasmoids on task bar, but they're happy to resize themselves. Removing "Task Manager" causes the system tray and digital clock to grow, which is disastrous when the task bar has vertical placement.
Default applications
1) No firefox?
Adept
1) Several seconds passed between the time the icon was clicked and the initial window was drawn.
2) Search results are not put in alphabetical order.
3) Two clicks are required to select a package for installation.
4) Very slow between clicking for installation and visual feedback.
5) The old behavior gave me an alphabetical list of every available package. This is good. I didn't have to guess how it was spelled or what tag is was filed under. How do I get this behavior back?
6) Searching for "git" gives no results. Searching for "git-" gives 11 results. Searching for "^git" gives >50 results, not all of which begin with "git". This is broken.
Konqueror
1) Spellchecking doesn't wait until the end of the word to decide that the word is misspelled. This results in annoying flashes of red underlining. No obvious way to turn off this behavior.
Home directory
1) What is screen-configurations.xml and why isn't it a dot file?
Taskbar
1) Abakus can be started from "Favorites" but when it is put on the task bar, kate executes.
Synaptic
1) Goes out of alphabetical order whenever I make a selection to install. |
|
|
| Gusty static IP issue |
[Aug. 28th, 2008|12:52 am] |
|
When using the GUI to change the IP address to static, "auto etho" is not added to /etc/network/interfaces. This prevents /etc/init.d/networking restart from bringing up the interface. Adding the line manually fixes the problem. |
|
|
| Making KDE 3x bearable |
[Aug. 20th, 2008|12:09 am] |
Cute animation when clicking on icons on the panel is annoying. Fix by kcontrol->peripherals->mouse->VisualFeedbackOnActivation=0
Bouncing busy cursor is annoying. Fix by kcontrol->Appearance...->BusyCursor=NoBusyCursor. |
|
|
| Intel compiler notes |
[Jan. 6th, 2007|07:07 pm] |
"feupdateenv is not implemented and will always fail" can be gotten rid of with
-i_dynamic |
|
|
| Potential programming assignment |
[Jan. 6th, 2007|03:17 pm] |
Peak Finding Algorithm.
Your program will take as input a data file with n space-separated columns. The first row will have labels for each of the columns, every remaining row will have a floating-point number in each column. The first n-1 columns are independent variables, the last column is a dependent variable.
For example, if there are three columns, they might look something like this:
a b c
1.0 0.0 3.2
1.0 0.5 3.3
1.0 1.0 3.4
0.0 0.0 2.3
0.0 0.5 2.5
0.0.1.0 2.7
The output of your program will be all of the local peaks in the parameter space. A local peak is defined as a point in parameter space that is higher than all of it's neighbors that are m steps away. A step is defined by values used by the parameters: for parameter a the step value is 1.0, and parameter b has a step value of .5.
The output file should retain the formatting of the input file. The correct output of the above example is:
a b c
1.0 1.0 3.4
Your program MUST handle the following case:
a) There are two independent parameters and one dependent parameter.
b) The values of the independent parameters range from 0.0 to 100.00.
c) The step value for each independent parameter is 1.0.
d) A local peak is determined by only looking at adjacent values that are a single step away.
e) The program should read the file from standard input and output results to standard output.
f) Values for the dependent variable are unique.
Your program SHOULD handle the following situations:
a) There are an arbitrary number of independent parameters.
b) The values of the independent parameters are arbitrary, but form a complete set of all possible combinations.
c) The step value for each independent parameter is arbitrary but consistent within the parameter (different parameters may differ).
d) The number of steps to search for the local peak is specified at runtime in the form of "-s ". Only Manhattan distance needs to be considered.
Your program MAY be tweaked to handle the following conditions:
a) Not all possible combinations of parameters are present.
b) "Local" may be defined in terms of Euclidean distance. Use "-e <float" on the command line.
c) "Local" may be defined in terms of both Euclidean and Manhattan distance. If both -e and -s are used
on the command line, the default is to combine them with a logical OR. If a "-a" is also on the command line, they will be combined with a logical AND (only points within the Manhattan distance AND the Euclidean distance are considered).
You may PROPOSE how to handle the situation where values for the dependent variable are not necessarily unique. |
|
|
| vim folding |
[Dec. 26th, 2006|11:21 am] |
:set foldmethod=indent
zo to open a fold
zc to close a fold
Decent writeup here. |
|
|
| L2 address clashes |
[Dec. 16th, 2006|10:03 pm] |
On opterons, L1 is index by virtual addresses, L2 by physical addresses. Physical addresses are not necessarily (and usually aren't) contiguous. And they'll change each time the program is rerun. If there's a physical address clash (or a sufficient number of them -- not sure if associativity mitigates this or not) then there will be an L2DCM for every access on those two (?) frames. Which will result in markedly different runtimes.
It's possible to mitigate this by reserving a chunk of contiguous physical memory at boot time and then.... well, doing a lot. And that would just guarantee consistent behavior, not necessarily optimal behavior.
Or, you can run your benchmark iterations out of the same process and get the same results for each run.
Or you can do multiple runs and pick the best value (or median value, or find the std dev, whatever).
Bleah. |
|
|
| OPM is dependent on IPC |
[Dec. 16th, 2006|09:57 pm] |
Well, CPI actually.
Ops per miss has an implicit assumption of how many cpu cycles there are per op. Change the cycles per op by going to a different instruction miss (say FDIV vs FADD) and the same OPM will have an order of magnitude greater CPU cycles, and thus a very different slowdown.
Thus, one could claim that each application has a specific OPM curve. However, application inputs may change the instruction mix as well as the opm range. |
|
|
| Cisco VPN Silliness |
[Nov. 9th, 2006|01:08 am] |
As per harisund
In the file linuxcniapi.c, there are 2 references to skb->stamp (one in line 452, one in line 315). Now in the Linux Kernel 2.6 this has been changed to skb->tstamp. This needs to be fixed in order to get rid of the compilation errors.
That a variable name changed in the kernel isn't silly. Not at all. The fact that I found this on Ubuntu's site, not Cisco's, is moderately silly. The fact that Cisco wanted me to fill out four pages of registration information (can you spot the invalid character in 'rountree'? Their software could!) to get a username and password that still wouldn't let me download a software update -- if one exists -- is very silly. Clue they have not.
Using vpnclient-x86_64-4.6.03.0190-k9. |
|
|
| (no subject) |
[Oct. 14th, 2006|05:04 am] |
root@malaise:/etc/init.d# cat /proc/cpuinfo
...
model name : AMD Athlon(tm) 64 Processor 3000+
...
And the homepage of blogspot can still bring it to its knees.
Nobody looks smarter wearing a baseball cap. No web site is
improved by animation.
</rant> |
|
|
| python comments in vim |
[Sep. 27th, 2006|01:17 pm] |
Python comments should not universally start at column 0. This fixes it.
autocmd BufRead *.py set smartindent
inoremap # X<c-h>#
Via Stüvel, thanks! |
|
|
| fish |
[Sep. 26th, 2006|03:41 pm] |
Really should get NFS installed.
Yup.
Really silly to have my email nicely isolated on a laptop, yet have to scp stuff to and from every other machine to use it.
Yup.
And really need to get a CUPS server set up, so I don't have to scp stuff from every machine to the machine that has the laser printer attached to it.
Yep, that needs to get done.
Oh, wait, I can just use the fish protocol.
fish://rountree@malaise
I know of no other profession that so consistently rewards abject laziness. |
|
|
| (no subject) |
[Sep. 25th, 2006|11:41 pm] |
For whatever reason, acroread is not in the repositories I'm using. So I finally broke down and installed it from the tarball on Adobe's site. Kept failing on a gtk library -- strace showed it finding the lib, opening it successfully, reading 512 bytes, closing it, and continuing to search. Hmmmmm.....
Ah, I need the 32-bit version.
apt-get install ia32-libs-gtk
Now it's complaining that a plugin isn't working -- oh well. I see that there's a 64-bit version in Edgy, so hopefully this won't happen again. |
|
|
| Tor on ubuntu dapper 6.06 |
[Sep. 21st, 2006|09:06 pm] |
I submitted this entry to the ora errata page and they finally got around to adding it to their "unconfirmed" list.
Looks like lots of other good stuff there, too. Go visit.
{279} Figure 7-8;
While I have had success with this hack, during a fresh install today I found the
following error in /var/log/tor/log:
"Aug 11 17:13:43.261 [warn] You are running Tor version 0.1.0.16, which will not work
with this network. Please use one of 0.1.0.18,0.1.1.23."
Following the Tor documentation to
http://wiki.noreply.org/noreply/TheOnionRouter/TorOnDebian , I installed Tor version
0.1.1.23 .
This brings me to figure 7-8. The figure is correct, but the dialog box is
sufficiently confusing that I think a bit more explanation would be helpful. "SOCKS
v4" and "SOCKS v5" have radio buttons, but neither are initially checked and it's not
clear if they refer to the text box "SOCKS Proxy" immediately above them (if the text
box is empty, shouldn't they be grayed out?). Unfortunately, if either is accidently
checked, there's no way to go back to the state of both being unchecked.
If one had to guess, "SOCKS v4" looks kinda sorta like the "forward-socks4a" that was
put into the config file. Unfortunately, that didn't work. Leaving both of them
unchecked or checking "SOCKS v5" does work for me.
While this GUI is (of course) not the fault of the authors, I think it's sufficiently
confusing to deserve a warning in the text. |
|
|
| Installing intel compilers |
[Sep. 18th, 2006|08:45 pm] |
|
This is stupid. Got all kinds of errors, warnings, and failures trying to install via sudo. Worked perfectly when I chose "ignore" at the first prompt and installed in my home directory. (Well, I had to create a link to the license file, but that was trivial.) |
|
|
| navigation |
| [ |
viewing |
| |
most recent entries |
] |
| [ |
go |
| |
earlier |
] |
| |
|
|
![[info]](http://l-stat.livejournal.com/img/userinfo.gif){kind=small}
barry_rountree's journal