PracticalWeb Ltd - websites that work for you

verbose remote ssh (with echo)

sean — Fri, 20 Jan 2012 16:47:31 +0000

I usually manage website deployment with bash scripts that run remote commands on the servers

In order that I can see what is going on and debug any errors verbose output is useful.

Running "bash -ex" causes each line to be output as it progresses and halt on any error so that you don't miss it.

within the remote command "set -x" cuase bash to echo all commands

#!/bin/bash -ex
ssh -T user@server << EOF
set -x
echo foo > /dev/null
EOF

Export from Git for Drupal site releases

sean — Tue, 15 Nov 2011 11:14:55 +0000

I'm used to how svn export works, and have used this in a number of release scripts.

Git doesn't quite have an equivalent

There is

git archive

Which creates a tarball rather than a set of files, and while you can export a subset of the repository you still get the full path in the tarball.

and

git checkout-index

Which creates a set of files (these can be locate outside the working copy) but requires a local repository.

Also I'm using github which doesn't support the remote repository option of git archive and thereadymade tarball doesn't seem to work via wget on a private project.

What I really wanted was a script I could hand to the sysadmin, that he could run on any server without needing a local git repository.

In the end I have instead written a script that packages the tarball I want and uploads that to a server he can access it from.

<?php
#!/bin/bash -ex

# pass a valid release tag as a parameter to this script to create an release tarball and upload to skunkworks

# this works on the local git repo - which must be up to date!
 
git archive   $1 www > /tmp/${1}.tar
cd /tmp
tar -xf ${1}.tar
rm ${1}.tar
rm -Rf www/sites/default/
mv www $1
tar -czf ${1}.tar.gz $1 
rm -Rf $1

scp ${1}.tar.gz dev.server.foo:/var/www/releases
?>

Then there is another script that downloads and unpacks this to upgrade our site.

<?php
#!/bin/bash -ex

# parameter is release tag

# tag has been released and uploaded to skunkworks

# live site is backed up

# live site docroot is /var/www/public_html

# each release is at /var/www/$release

# symlink points /var/www/public_html to current release

# sites/default is common to every release - stored as /var/www/sites-default
# symlink added to each release to point to this

export COLUMNS=80

release=$1

cd /var/www/

wget http://dev.server.foo/releases/${release}.tar.gz

tar -xzf ${release}.tar.gz

cd ${release}

ln -s /var/www/sites-default sites/default

drush -y vset site_offline_message "<p>Site offline for upgrade</p>" 
drush -y vset site_offline 1 

drush -y updb

drush cc all

# if the current user is a member of the www-data or apache group we can make the files owned by this group
# as long as apache has umask 002 files should now be writable by us and apache
groups | grep www-data > /dev/null && find sites/default/files/ ! -group www-data -exec chgrp www-data {} ;
groups | grep apache > /dev/null && find sites/default/files/ ! -group apache -exec chgrp apache {} ;

cd ..

rm public_html

ln -s ${release} public_html

cd ${release}

drush -y --uri=www.defaqto.com vset site_offline 0
?>

Instance parameter for drupal field_create_instance($instance)

sean — Fri, 23 Sep 2011 17:30:14 +0100

The code in chapter 7 of pro Drupal development has a bug.

The $instance array that you need to pass to field_create_instance() needs to be as below.

Note that the display array has the keyword, 'default' for the default display - ie a node on it's own page 'teaser' looks like the right value for teaser. Other view modes are possible too.

The label field in the display part of the array is just the positioning of the array - 'above', 'below', 'inline' and 'hidden' seem the likely set of allowed values.

NB the display part of the array is optional as Drupal will provide default values.

<?php
array( 
    array(
        'field_name' => 'job_post_company',
        'type'       => 'text',
        'label'      => $t('Company posting the job listing'),
        'widget'     => array(
                   'type' => 'text_textfield',
         ),
        'display' => array(
           'default' => array(
               'label' => 'above',
               'type'  => 'text',
         ),
     ),
   ),
);
?>

The errata for this book seems to be long and unhelpfully laid out 563 reported errors so far...

http://www.drupalbook.com/errata3

There even seems to be errors in the errata http://www.drupalbook.com/node/269 seems to report this error (or part of it) but I think it's wrong - the array key should be a view mode.

node_save tags by term name with auto addition of new tags

sean — Thu, 22 Sep 2011 18:12:28 +0100

Drupal lets you add a taxonomy term to a node and if you use the GUI you can have a nice auto-complete widget that takes term name - and iof the term doesn't exist yet you get a new term created when you save the node.

I wanted to access the same functionality on node I am generating from external data.

The key is to create the term array with tid = 'autocreate' - this and the vocabulary ID are enough to trigger the taxonomy module to lookup the tid for you - and adda new term is required.

<?php
$node = new stdClass();
$node->type='mytype';
$node-title='foo';
$node->body = 'bar';

$node->field_cat_tag['und'][0]['name'] =  $name;
$node->field_cat_tag['und'][0]['vocabulary_machine_name'] = $vocab_machine_name; // optional - helps performance if in a loop  
$node->field_g4_cat_tag['und'][0]['vid'] = $vid;
$node->field_g4_cat_tag['und'][0]['tid'] = 'autocreate';

$node_save($node);
?>

Using PHP and curl with http PUT on string data

sean — Wed, 21 Sep 2011 15:59:40 +0100

This wasn't obvious

There are various other posts on this - but mostly assuming you will put a file.

In my case I want to put the contents of a string so I needed to craete a temporary filehandle.

<?php
function put_it($url, $string) {
  $ch = curl_init($url);
  curl_setopt($ch, CURLOPT_RETURNTRANSFER, true);
  curl_setopt($ch, CURLOPT_PUT, TRUE);
  // create tempoary file handle
  $oneMB = 1024 * 1024;
  $fp = fopen("php://temp/maxmemory:$oneMB", 'r+');
  fputs($fp, $string);
  rewind($fp);
  curl_setopt($ch, CURLOPT_INFILE,  $fp);
  curl_setopt($ch, CURLOPT_INFILESIZE, strlen($string));
  $response = curl_exec($ch);
  fclose($fp);
?>

RAID ext4 disk problems

sean — Tue, 20 Sep 2011 14:20:45 +0100

One of those posts for myself in case I need to come back to this later

I have a new PC which is very fast apart from problems with disc access seeming slow (and the disks being physically noisy)

The following related links led me to try a solution

http://forums.gentoo.org/viewtopic-t-843292.html

http://marc.info/?l=linux-raid&m=128506852210452&w=2

/etc/fstab
UUID=317f7912-9956-41f1-9855-7bee69c950b3 / ext4 errors=remount-ro,barrier=0 0 1

NB adding barrier=0 here seems to have solved the problem

How to turn off notifications in Banshee

sean — Fri, 19 Aug 2011 11:21:51 +0100

The music player banshee which is part of the default Ubuntu setup pops up a message each time a new track starts playing.

To turn this off

Open Banshee
Go to Edit, preferences, extensions
Enable Notification Area Icon
The Icon should now show up - right click on it
Uncheck Show notifications
Disable the Notification Area Icon again

Very annoying - the option to turn off notifications is hidden on a menu you can't get to by default.

Henri Bergius' secret agenda for PHP Content Management Systems

sean — Wed, 13 Jul 2011 09:31:23 +0100

I read this post the other day and found it really inspiring - so many things to do and learn...

http://bergie.iki.fi/blog/my_secret_agenda_for_php_content_management_sy...

Links within that blog that I found especially interesting

https://github.com/indeyets/appserver-in-php

http://documentcloud.github.com/backbone/

https://github.com/bergie/VIE

http://phpcr.github.com/

svn merges and excess merginfo properties

sean — Wed, 13 Jul 2011 09:20:08 +0100

The main svn server I'm working on has finally upgraded to svn 1.5 - so easier merging is now possible - hooray.

But we've been seeing merges that should only affect one file resulting in property changes on lots of files/directories. This isn't especially harmful - but it is annoying and makes reviewing the merge harder because of the extra noise.

It turns out that this extra mergeinfo is likely because of earlier merges on subtrees http://blogs.collab.net/subversion/2009/11/where-did-that-mergeinfo-come...

It is possible to remove the subtree mergeinfo via this command run from the root of the branch

svn propdel -R svn:mergeinfo ./*

http://stackoverflow.com/questions/767418/remove-unnecessary-svnmergeinf...

Which should get rid of the problem - I'm just not 100% sure if it throws away important data.

For now I think I'll live with the nuisance - and come back to this if it becomes too annoying or if I gain greater insight into the issue.

Selenium test for computedStyle

sean — Tue, 28 Jun 2011 08:41:40 +0100

I wanted to write some automated tests for CSS changes and bugs, these type of issues are quite prone to regression so re-running the tests has a large benefit.

It is possible to add new assertions to selenium via the user-extensions.js file

The following adds a test for computed style, this only works in firefox as far as I know - but some automation is better than none

//only works in firefox
Selenium.prototype.assertStyle = function(locator, text) { 
    var params = text.split('=', 2);
    
    var propertyName = params[0];
    var expectedValue = params[1];
    
    var element = this.page().findElement(locator);
    var actualValue = this.browserbot.getCurrentWindow().getComputedStyle(element, null).getPropertyValue(propertyName);
    Assert.matches(expectedValue, actualValue);
}; 

Now I can add a selenium test like

<table>
<tr>
	<td>assertStyle</td>
	<td>//td[@class='lender'][1]</td>
	<td>text-align=center</td>
</tr>

</table>

Selenium only lets me pass two parameters and really I want three (locator, css property, expected value). So I've hacked this and put two parameters into one - which is a bit smelly but for now gets the job done - comments on better methods very welcome.

I need to at least check that the 'text' parameter is validated properly - but I figured I'd blog now and update later as I've a deadline to meet today.