Tinypliny's Journal
My Podcast Link
10/14/2011 22:19 #55298
Gadgety GoodnessCategory: i-tech
10/14/2011 18:30 #55297
Fedora missing fonts for Gnumeric?Category: linux
10/14/2011 10:14 #55295
Suse Suse Studio?Category: linux
Though Debian has been my favourite in terms of usability and friendliness, there is something really wrong with how Debian systems interact with R and deal with numerical ordering.
For eg. you would expect intervals to be ordered as:
[ 7,11)
[11,12)
[12,13)
[13,14)
[14,29]
Instead Debian orders them as:
[11,12)
[12,13)
[13,14)
[14,29]
[ 7,11)
In a scenario where I have one-digit, two-digit and three-digit numbers, I get the weirdest possible ordering. In the example above, it compares 1 and 7, rather than 11 and 7. This number-by-number evaluation creates serious problems in trend testing and determination of direction of association between variables. It potentially distorts the category that is treated as the reference variable. I looked at R documentation to see if I could change something to make it right. However, after many trials, my windows machine and machines running Fedora were uniformly giving me the right answers while Debian derivatives were giving me the weird answers.
For now, this is a dealbreaker for me and I can't go back to Debian.
That leaves me with
-- CentOS
-- Scientific Linux
-- Fedora
-- OpenSuse
I tried installing CentOS and Scientific Linux on the brawny Toshiba this past weekend but first there were problems with permissions and then problems with the R repository in both CentOS and Scientific Linux. Though Scientific Linux claims to come with R, the DVD I downloaded did not have R! When I tried to install R, there was no R in the default repository. Adding the correct R repository was an unending nightmare that took up 2 hours. I gave up in frustration. Then there was a problem with permissions. CentOS and Scientific Linux don't set up a username to go with the root password in the installation sequence. Even though you might have given a name to your computer during installation, you start up with something like:
yourgivenname@localhost.
instead of
yourgivenname@yourgivencomputername
In addition, it doesn't automatically add you to the sudoers group. I know linux admins won't view this as a problem but end-users like me don't need to work out details about how to add themselves to sudoers, change the localhost name, set up root permissions once more after installation. Its a pain to be burdened with all these nitty grittys even before you start using the system.
CentOS and Scientific Linux are so similar in their installation setup and behaviour, they might as well be clones of each other. The permissions and repositories problem rears its ugly head at every possible software installation. It gets to be very tiring. And what is with the repositories anyway??! They have software from 2 years ago! I understand they are more stable than the newer ones, but come on, newer features that make life easier in the software I use are very important to me!
So no more RHEL derivatives for me.
Which leaves me with Fedora. Sigh. I don't know if I can deal with all the mess that I have been documenting.
So at this point, I am thinking Suse Studio with the software I need from the list below. It could be better than getting OpenSuse and then installing everything because adding software and repositories beforehand might reduce dependency conflicts.
What's the point of this sort of ordering??! It so illogical for so many reasons!
github is doing the same stupid alpha sort to my numeric tags. :::link::: Why couldn't they just put them in reverse tstamp order, or natural order.
Reading your comments, I really don't think I am "linux-ready" yet. LOL I feel like I jumped into a war with no weapons!
Not being in the wheel group by default is trivial to resolve using usermod or the GUI aand allowing wheel users to sudo is resolved by uncommenting the line at the bottom of /etc/sudoers where it enables wheel users to sudo.
I think thr idea is if you can't do those basic things in Linux you have no business running a server OS.
Open suse =novella=in bed with Microsoft. They used to be my favorite too.
Fedora is better. The issue you are experiencing with scientific Linux and cebtos is that they are really designed to be stable servers not desktops. They are both clones of RHEL. The whole idea of those stable OSes is that you can build and base code on stable versions so that if you were writing something that did a years worth of analysis you don't have to worry about the libraries and APIs changing every week.
Why don't you come by and we can set you up on fedora OR Scientific Linux. There is nothing stopping you from using more current libraries on scientific Linux.
The debian problem cannot be a debian problem but a problem with Rs port on debian.
10/14/2011 05:34 #55294
What do I need?Category: linux
Analyses Environment
- R
- RStudio
- Variable number of R packages
Spreadsheet
- Gnumeric
Writing Environment
- Texlive
-
- Texworks
-
- Geany
-
- WriteType
- Zim desktop wiki
- LibreOffice Writer.
Networking
- Wireless
- Wired
- (Bluetooth would be nice, but seldom works, so I could let it go.)
Media
-
Graphics
- Gimp
- Mypaint
-
... to be continued as I think of more things I use.
EDIT: Turns out I need WAAY more things that this puny list, so I decided to attached my latest yum log as a text file here.
Type:
sudo geany /var/log/yum.log
And you will get this log with date and time details of every package you have ever updated, installed or erased in Fedora 15.
[txt]1011/yum102311.txt[/txt]
This should all work on fedora.
10/14/2011 03:55 #55293
Worst luck indeed: Fedora 15 BSODCategory: linux
Transcript of Error:
[5.693915] nouveau 0000:01:00.0: Invalid ROM contents
[5.825510] nouveau 0000:01:00.0: Pointer to BIT loadval table invalid
[5.849874] nouveau 0000:01:00.0: 0XD518: i2c wr fail: -6
[5.964968] nouveau 0000:01:00.0: PGRAPH: unsupported chipset, please report!
[6.092104] [drm: intel_dsm_platform_mux_info] *ERROR* MUX INFO call failed
Looks like I might need a hard reboot. The wretched thing is not coming out of this error screen. And it goes into it at every boot and suspend.
Apparently, this is a known new bug:
With some insights from here:
ds,
Your issue is related to the optimus technology deployed in the m14x, it's a new piece of tech that allows switching between your nvidia card and intel sandy bridge. In short, your nvidia card does not work on Linux out-of-the-box. Read more about how you can either disable your nvidia card (I did that and saved 14 w (1½ hours of battery time) or try and enable it using some experimental tech called switcheroo. Read more here:
Posted by Magnus Glantz 2011-07-06 01:07:30
My brother's hacky solution:
Close the lid for 15-30 seconds and then open. If it doesn't work, repeat as many times as necessary.
I can only say... !!!
Yes, there is a sandy bridge. The laptop has onboard intel graphics AND discrete nvidia graphics card with a kind of hybrid switching technology. The problem is this kind of switching is not supported in linux so it throws errors like the one above.
Now I have no way to adjust resolution or brightness. There is no way for me to adjust/turn it off in BIOS. I checked.
The usual way to turn graphic cards on and off is in the bios.
My assumption that the nvidia card is an add-on (probably pci express) and the laptop came with a main board with sandy bridge bridge providing onboard graphics. But as always, I could be wrong.
The error has gone away after a fresh install of F15 with xfce. But I need to figure out how to turn the NVIDIA card off if it turns up again in the future. Both the solutions from here: :::link::: acpi_call and byo-switcheroo don't work for Toshiba P750. I need to touch base with (e:Paul) about this sometime this week and post an update.
Plagued by this once more. I am seriously doomed.
Stoke the fires, why don't you.
Its yours, mine looks totally normal.