I bought a TonidoPlug, which is a wall-wart with an embedded Linux server (running Ubuntu 9.04 Jaunty). Out of the box, sudo gave the error:
sudo: must be setuid root
This message means exactly what it says -- the sudo command is not setuid and it must be setuid (and owned by root).
chown root:root /usr/bin/sudo chmod u+s /usr/bin/sudo
will correct the 'problem.' (I put quotes around 'problem' because there are some who will say that this is a security feature.)
There are hundreds of reports on the net about "sudo: gem: command not found". Out of the dozen or so I looked at, the suggested resolution amounted to "ensure that ‘gem’ is in your PATH."
Yep. It is in my path. That is to say, the following works:
gem –help
but the following fails:
sudo gem –help
Further confusing matters, gem is in the path shown by:
sudo echo $PATH
The trick is that sudo doesn’t use $PATH for its path on some editions of Linux. This behavior is considered "more secure." (See http://stackoverflow.com/questions/257616/sudo-changes-path-why ).
Bah! It is still a pain to work around. You have to get ‘gem’ into the path (not simply the $PATH!) You’ll read about options you can pass to sudo, and that is fine and dandy when you type the command-line yourself. When someone’s script invokes sudo for you, you’re stuck with the command line that it uses.
The hack, reproduced from the stackoverflow article:
mv /usr/bin/sudo /usr/bin/sudo.orig
cat > /usr/bin/sudo <<EOF
#!/bin/bash
/usr/bin/sudo.orig env PATH=$PATH "$@"
EOF
then your regular sudo works just like the non secure-path sudo.
Here is a side-by-side of Bort and Suspenders. Both are Rails starter applications.
September 10, 2009
| Opinion | Bort | Suspenders |
| pagination | mislav-will_paginate | mislav-will_paginate |
| non-HTML mark-up |
|
RedCloth 4 |
| mocking |
|
mocha |
| unit test | rspec, rspec_rails | shoulda |
| unit test "noise suppression" |
|
quietbacktrace |
| Error notification | exception notifier | hoptoad_notifier |
| forms handler | Fudge Form | formtastic |
| file attachment | attachment fu OR Paperclip | paperclip |
|
|
|
helper_test |
| fixtures |
|
factory_girl |
| deployment | capistrano | capistrano |
| web server |
|
mongrel cluster |
| VCS |
|
git |
| time mocking |
|
Timecop |
| login | Restful Authentication (includes OpenID) | Clearance |
| integration testing
|
|
Cucumber |
| Browser simulator |
|
Webrat |
| Extra rake tasks | bort:install (installs optional add-ons) | Limerick Rake |
| Javascript smoosher | asset packager OR bundle fu* |
|
| roles | Role Requirement *
|
|
| tagging | acts_as_taggable_on_steroids * |
|
| alternative to RHTML | HAML *
|
|
Note that in the above table, more opinions is neither good nor bad -- unless they match/don't-match your opinions.
* = optional via "rake bort:install"
For myself, I like the Thoughtbot (Suspenders) stack with the following exceptions:
Bundling in Cucumber, Webrat, and Clearance is nice. They fit well together.
So here’s my current rails project start-up:
So I created the attached shell script to automate this.
UPDATE (9/16/2009): I’ve abandoned this approach. Too often I work from inside a restrictive firewall, and the gyrations necessary to pull-in all of the pieces are simply too idiosyncratic to automate reliably.
If you get
Anonymous modules have no name to be referenced by
when you’re using a Rails template, it can be caused by referencing an undefined variable. You can pretend that message means that the [nothing.] in [nothing.]VARIABLE_NAME is an anonymous module.
I wanted to add a project to Subversion, but I wanted to ignore some of the directories in the project directory. The standard way to do this would be to add the whole project directory and then set the svn:ignore property to ignore the directories, and then delete them from the repository. But that adds the directories to svn, and delete doesn’t really delete, so I wanted better.
I was getting "can’t activate , already activated nokogiri-1.3.3" for some environments, but not others. In particular, I got it when I ran:
rake environment RAILS_ENV=test db:migrate
I’m using Suspenders, and config/test.rb and config/cucumber.rb specify version 1.3.2. Solution: Remove the version specification from these files.
Webrat is a lovely testing tool, particularly when used with Cucumber. But every time I put it aside for a few weeks, when I come back to it, I forget about this…
And I fill in "name" with "John Doe"
Could not find field: "name" (Webrat::NotFoundError)
but when I go to the .html.erb file, I see the field, plainly defined:
<%= f.text_field :name %>
So why doesn’t webrat find the field? Because the field is NOT named "name". It is named "user_name". This is because it is wrapped inside
<% form_for :user, :url => users_path do |f| -%>
Because of the form_for, all field names get prefixed with the form name.
I had a strange problem with Firefox 3.5.2 and Google Desktop. Searching GDS via double-control, would send Firefox to an URL resembling:
http://127.0.0.1:4664/search?s=1234567894&q=searchterm&aq=t
which would redirect to my external IP address, and then fail (because I’m not running a web server).
When I switched from Firefox as my default browser to Internet Explorer, double-control search worked. Furthermore, Firefox would work (i.e. not redirect) if I pasted
http://localhost:4664/search?s=1234567894&q=searchterm&aq=t
into the address bar manually.
I tried disabling add-ons, turning off my proxy.pac file, and starting in safe mode. No joy. However, clearing the cache did fix the problem.
From time to time you may read that Intuit needlessly dropped QIF import from Quicken. Intuit has stated that QIF was dropped because it wasn’t robust enough. I won’t speak to the strengths or weaknesses of QIF as a format, but I can tell you that Quicken 2006’s support for QIF is lame.
Others have documented that, despite Quicken’s claims to the contrary, you can import QIF files, so long as they have a "!Account" header added. This is true – to a limited extent.
I’ve got a Quicken file with over 22,000 transactions in it. Pretty much my entire financial history for about 30 years. (At one point, I went back and entered data from before Quicken existed. I had waaay too much time on my hands.) Because I had reason to suspect that there was some corruption in the data file that super-validating wasn’t fixing, I decided to "clean up" my data by exporting the entire file to QIF, creating a new Quicken data file, and importing the QIF back in.
After exporting, I checked the QIF file, and it looked good. After importing the first time, the results were terrible. One of the problems I had was incomplete import of split transactions, and another was that import had somehow renamed all my utility bills to "My Bank Home Mortgage". So I created a new file, and turned off transaction renaming, and I imported my accounts with many "split" transactions first.
This worked better, for some small value of better. Instead of being off by a million dollars over 30 years, I was off by only $100,000.
I spent some time comparing the before-export and the after-import files, and comparing them to the QIF file. When the QIF file has a transaction with many split items, and some of the later splits are transfers to other accounts, they aren’t imported.
Ouch!