Java 2 Ada

New debian repository with Ada packages

By Stephane Carrez 2014-04-06 15:54:30

I've created and setup a Debian repository to give access to several Debian packages for several Ada projects that I manage. The goal is to provide some easy and ready to use packages to simplify and help in the installation of various Ada libraries. The Debian repository includes the binary and development packages for Ada Utility Library, Ada EL, Ada Security, and Ada Server Faces.

Access to the repository

The repository packages are signed with PGP. To get the verification key and setup the apt-get tool, you should run the following command:

wget -O - http://apt.vacs.fr/apt.vacs.fr.gpg.key | sudo apt-key add -

Ubuntu 13.04 Raring

A first repository provides Debian packages targeted at Ubuntu 13.04 raring. They are built with the gnat-4.6 package and depend on libaws-2.10.2-4 and libxmlada4.1-dev. Add the following line to your /etc/apt/sources.list configuration:

deb http://apt.vacs.fr/ubuntu-raring raring main

Ubuntu 12.04 LTS Precise

A second repository contains the Debian packages for Ubuntu 12.04 precise. They are built with the gnat-4.6 package and depend on libaws-2.10.2-1and libxmlada4.1-dev. Add the following line to your /etc/apt/sources.list configuration:

deb http://apt.vacs.fr/ubuntu-precise precise main

Installation

Once you've added the configuration line, you can install the packages:

sudo apt-get update
sudo apt-get install libada-asf1.0

For the curious, you may browse the repository here.

Ada Server Faces 1.0.0 is available

By Stephane Carrez 2014-04-05 19:05:21

Ada Server Faces is a framework that allows to create Web applications using the same design patterns as the Java Server Faces (See JSR 252, JSR 314, or JSR 344). The presentation pages benefit from the Facelets Web template system and the runtime takes advantages of the Ada language safety and performance.

A new release is available with several features that help writing online applications:

  • Add support for Facebook and Google+ login
  • Javascript support for popup and editable fields
  • Added support to enable/disable mouseover effect in lists
  • New EL function util:iso8601
  • New component <w:autocomplete> for input text with autocompletion
  • New component <w:gravatar> to render a gravatar image
  • New component <w:like> to render a Facebook, Twitter or Google+ like button
  • New component <w:panel> to provide collapsible div panels
  • New components <w:tabView> and <w:tab> for tabs display
  • New component <w:accordion> to display accordion tabs
  • Add support for JSF <f:facet>, <f:convertDateTime>, <h:doctype>
  • Support for the creation of Debian packages

You can try the online demonstration of the new widget components and download this new release at http://download.vacs.fr/ada-asf/ada-asf-1.0.0.tar.gz

Ada Security 1.1.0 is available

By Stephane Carrez 2014-03-22 14:05:36

The Ada Security library provides a security framework which allows applications to define and enforce security policies. This framework allows users to authenticate by using OpenID Authentication 2.0, OAuth 2.0 or OpenID Connect protocols.

The new version brings the following improvements:

  • New authentication framework that supports OpenID, OpenID Connect, OAuth, Facebook login
  • AWS demo for a Google, Yahoo!, Facebook, Google+ authentication
  • Support to extract JSON Web Token (JWT)
  • Support for the creation of Debian packages

The library can be downloaded at http://download.vacs.fr/ada-security/ada-security-1.1.0.tar.gz

Ada EL 1.5.0 is available

By Stephane Carrez 2014-03-19 21:49:33

Ada EL is a library that implements an expression language similar to JSP and JSF Unified Expression Languages (EL). The expression language is the foundation used by Java Server Faces and Ada Server Faces to make the necessary binding between presentation pages in XML/HTML and the application code written in Java or Ada.

The presentation page uses an UEL expression to retrieve the value provided by some application object (Java or Ada). In the following expression:

#{questionInfo.question.rating}

the EL runtime will first retrieve the object registered under the name questionInfo and look for the question and then rating data members. The data value is then converted to a string.

The new release is available for download at http://download.vacs.fr/ada-el/ada-el-1.5.0.tar.gz

This version brings the following improvements:

  • EL parser optimization (20% to 30% speed up)
  • Support for the creation of Debian packages

Ada Utility Library 1.7.0 is available

By Stephane Carrez 2014-02-09 11:06:03

Ada Utility Library is a collection of utility packages for Ada 2005. A new version is available which provides:

  • Added a text and string builder
  • Added date helper operations to get the start of day, week or month time
  • Support XmlAda 2013
  • Added Objects.Datasets to provide list beans (lists of row/column objects)
  • Added support for shared library loading
  • Support for the creation of Debian packages
  • Update Ahven integration to 2.3
  • New option -r <test> option for the unit test harness to execute a single test
  • Port on FreeBSD

It has been compiled and ported on Linux, Windows, Netbsd, FreeBSD (gcc 4.6, GNAT 2013, gcc 4.7.3). You can download this new version at http://download.vacs.fr/ada-util/ada-util-1.7.0.tar.gz.

Migrating a virtual machine from one server to another

By Stephane Carrez 2014-01-24 23:05:37

OVH is providing new offers that are cheaper and provide more CPU power so it was time for me to migrate and pick another server and reduce the cost by 30%. I'm using 7 virtual machines that run either NetBSD, OpenBSD, FreeBSD or Debian. Most are Intel based, but some of them are Sparc or Arm virtual machines. I've colllected below the main steps that must be done for the migration.

LVM volume creation on the new server

The first step is to create the LVM volume on the new server. The volume should have the same size as the original. The following command creates a 20G volume labeled netbsd.

$ sudo lvcreate -L 20G -Z n -n netbsd vg01
WARNING: "netbsd" not zeroed
Logical volume "netbsd created

Copying the VM image

After stopping the VM, we can copy the system image from one server to another server by using a combination of dd and ssh. The command must be executed as roototherwise some temporary file and additional copy steps could be necessary.

$ sudo dd if=/dev/vg01/netbsd bs=8192 |
ssh root@master.vacs.fr dd bs=8192 of=/dev/vg01/netbsd

root@master.vacs.fr's password:
2621440+0 records in
2621440+0 records out
21474836480 bytes (21 GB) copied, 1858.33 s, 11.6 MB/s
2621440+0 records in
2621440+0 records out
21474836480 bytes (21 GB) copied, 1848.62 s, 11.6 MB/s

By compressing the image on the fly, the remote copy is faster (4 times faster). The following command does this:

$ sudo dd if=/dev/vg01/netbsd bs=8192 |
gzip -c | ssh root@master.vacs.fr \
'gzip -c -d | dd bs=8192 of=/dev/vg01/netbsd'

root@master.vacs.fr's password:
2621440+0 records in
2621440+0 records out
21474836480 bytes (21 GB) copied, 427.313 s, 50.3 MB/s
2621440+0 records in
2621440+0 records out
21474836480 bytes (21 GB) copied, 436.128 s, 49.2 MB/s

Once the copy is done, it's good to verify the integrity of the copy. For this, we can run the sha1sumon the source image and on the destination image and compare the SHA1 checksum: they must match.

$ sudo sha1sum /dev/vg01/netbsd
04e23ccc1d22cb1de439b43535855b2d1331da6a /dev/vg01/netbsd

(run this command on both servers and compare the result).

Importing the virtual machine definition

The last step is to copy the virtual machine definition from one server to the other. The definition is an XML file located in the /etc/libvirt/qemu directory. Once copied, run the virsh command on the target server and import the definition:

$ sudo virsh
virsh# define netbsd.xml
virsh# start netbsd

That's it, the virtual machine was migrated at a reasonable small cost: the whole process took less than one hour!

Installation of FreeBSD for a jenkins build node

By Stephane Carrez 2013-12-31 10:59:44

A few days ago, I did a fresh installation of my Jenkins build environment for my Ada projects (this was necessary after a disk crash on my OVH server). I took this opportunity to setup a FreeBSD build node. This article is probably incomplete but tends to collect a number of tips for the installation.

Virtual machine setup

The FreeBSD build node is running within a QEMU virtual machine. The choice of the host turns out to be important since not all versions of QEMU are able to run a FreeBSD/NetBSD or OpenBSD system. There is a bug in QEMU PCI emulation that prevents the NetBSD network driver to recognize the emulated network cards (See qemu-kvm 1.0 breaks openbsd, netbsd, freebsd). Ubuntu 12.04 and 12.10 provide a version of Qemu that has the problem. This is solved in Ubuntu 13.04, so this is the host linux distribution that I've installed.

For the virtual machine disk, I've setup some LVM partition on the host as follows:

sudo lvcreate -Z n -L 20G -n freebsd vg01

this creates a disk volume of 20G and label it freebsd.

The next step is to download the FreeBSD Installation CD (I've installed the FreeBSD-10.0-RC2). To manage the virtual machines, one can use the virsh command but the virt-manager graphical front-end provides an easier setup.

sudo virt-manager

The virtual machine is configured with:

  • CPU: x86_64
  • Memory: 1048576
  • Disk type: raw, source: /dev/vg01/freebsd
  • Network card model: e1000
  • Boot on the CD image

After the virtual machine starts, the FreeBSD installation proceeds (it was so simple that I took no screenshot at all).

Post installation

After the FreeBSD system is installed, it is almost ready to be used. Some additional packages are added by using the pkg install command (which is very close to the Debian apt-get command).

pkg install jed
pkg install sudo bash tcpdump

By default the /proc is not setup and some application like the OpenJDK need to access it. Edit the file /etc/fstab and add the following lines:

fdesc   /dev/fd         fdescfs         rw      0       0
proc    /proc           procfs          rw      0       0

and mount the new partitions with:

mount -a

GNAT installation

The FreeBSD repository provides some packages for Ada development. They are easily installed as follows:

pkg install gmake
pkg install gnat-aux-20130412_1 gprbuild-20120510
pkg install xmlada-4.4.0.0_1 zip-ada-45
pkg install aws-3.1.0.0
pkg install gdb-7.6.1_1

After the installation, change the path and setup the ADA_PROJECT_PATH variables to be able to use gnatmake:

export PATH=/usr/local/gcc-aux/bin:$PATH
export ADA_PROJECT_PATH=/usr/local/lib/gnat

Jenkins slave node installation

Jenkins uses a Java application that runs on each build node. It is necessary to install some Java JRE. To use subversion on the build node, we must make sure to install some 1.6 version since the 1.8 and 1.7 version have incompatibilities with the Jenkins master. The following packages are necessary:

pkg install openjdk6-jre-b28_7
pkg install subversion-1.6.23_2

Jenkins needs a user to connect to the build node. The user is created by the adduser command. The Jenkins user does not need any privilege.

Jenkins master will use SSH to connect to the slave node. During the first connection, it installs the slave.jarfile which manages the launch of remote builds on the slave. For the SSH connection, the password authentication is possible but I've setup a public key authentication that I've setup on the FreeBSD node by using ssh-copy-id.

At this stage, the FreeBSD build node is ready to be added on the Jenkins master node (through the Jenkins UI Manage Jenkins/Manage Nodes).

MySQL Installation

The MySQL installation is necessary for some of my projects. This is easily done as follows:

pkg install mysql55-server-5.5.35 mysql55-client-5.5.35

Then add the following line to /etc/rc.conf

mysql_enable="YES"

and start the server manyally:

/usr/local/etc/rc.d/mysql-server onestart

The database tables are setup during the first start.

Other packages

Some packages that are necessary for some projets:

pkg install autoconf-2.69 curl-7.33.0_1
pkg install ImageMagick-nox11-6.8.0.7_3

Jenkins jobs

The jenkins master is now building 7 projects automatically for FreeBSD 10: FreeBSD Ada Jobs

Suivi de consommation éléctrique avec clef USB Teleinfo ADTEK

By Stephane Carrez 2013-09-22 08:57:37

Les compteurs EDF récent disposent d'un module émettant périodiquement des informations sur la consommation éléctrique. Le compteur utilise un protocol série à 1200 baud, le signal est modulé par une porteuse à 50Khz (Voir téléinformation EDF pour les détails ainsi que la Spéficiation Technique EDF). Cet article explique comment récupérer ces informations et les rendre visibles à travers plusieurs graphes. En deux mots, le principe est de récupérer les informations EDF, d'envoyer ces informations sur un serveur et afficher tous les graphes et résultats à travers un interface Web accessible depuis Internet.

bbox-teleinfo.png

Téléinformation avec clef USB ADTEK

La société Adtek propose un petit module Téléinfo USB permettant de récupérer la téléinformation via un port série. La communication se fait à 9600 baud, 8-bits, sans parité. Sous Linux, il faut charger les deux modules usbserialet ftdi_sio. Suivant la version du driver ftdi, la clef USB peut ne pas être reconnue, il faut alors indiquer les identifiants du fabricant et du produit lors du chargement du driver.

insmod usbserial.ko
insmod ftdi_sio.ko vendor=0x0403 product=0x6015

Si tout se passe bien le driver va créer le device /dev/ttyUSB0 lorsque la clef est montée:

usbserial: USB Serial Driver core
USB Serial support registered for FTDI USB Serial Device
ftdi_sio 2-2:1.0: FTDI USB Serial Device converter detected
usb 2-2: Detected FT232RL
usb 2-2: FTDI USB Serial Device converter now attached to ttyUSB0
usbcore: registered new interface driver ftdi_sio
ftdi_sio: v1.4.3:USB FTDI Serial Converters Driver

Petit agent de monitoring

Un petit agent de monitoring va lire en permanence les trames EDF de téléinformation via le port série. Il doit collecter les données et envoyer les résultats toutes les 5 minutes en utilisant un POST HTTP vers le serveur qui lui est donné au démarrage.

edf-teleinfo /dev/ttyUSB0 http://server/teleinfo.php &

Cet agent peut tourner dans un Raspberry Pi, un BeagleBone Black. Dans mon cas, je le fais tourner sur ma Bbox Sensation ADSL. A défaut, on peut utiliser un PC standard mais ce n'est pas optimal pour la consommation éléctrique. Source de l'agent: edf-teleinfo.c

La compilation de l'agent se fait simplement avec l'une des commandes suivantes:

gcc -o edf-teleinfo -Wall -O2 edf-teleinfo.c
arm-angstrom-linux-gnueabi-gcc -o edf-teleinfo-arm -Wall -O2 edf-teleinfo.c

Création des fichiers RRDtool

Le compteur EDF envoie une mesure toutes les 2 secondes (option -s de rrdtool). La consommation éléctrique est enregistrée sous deux data sources: hc (Heures creuses) et hp (Heures pleines). Les min, max et average sont calculés pour des périodes de 1 mn (30 mesures), 5mn (150 mesures) et 15 mn (450 mesures).

rrdtool create teleinfo-home.rrd -s 2 \
   DS:hc:COUNTER:300:0:4294967295 \
   DS:hp:COUNTER:300:0:4294967295 \
   RRA:AVERAGE:0.1:30:1800 \
   RRA:MIN:0.1:30:1800 \
   RRA:MAX:0.1:30:1800 \
   RRA:AVERAGE:0.1:150:1800 \
   RRA:MIN:0.1:150:1800 \
   RRA:MAX:0.1:150:1800 \
   RRA:AVERAGE:0.1:450:1800 \
   RRA:MIN:0.1:450:1800 \
   RRA:MAX:0.1:150:1800

Alors que les Heures creuses et Heures pleines sont définies comme COUNTER, l'intensité instantanée et la puissance apparente sont représentées avec des gauges variant de 0 à 70A ou 0 à 15000W.

rrdtool create teleinfo_power-home.rrd -s 2 \
   DS:ic:GAUGE:300:0:70 \
   DS:pap:GAUGE:300:0:15000 \
   RRA:AVERAGE:0.1:30:1800 \
   RRA:MIN:0.1:30:1800 \
   RRA:MAX:0.1:30:1800 \
   RRA:AVERAGE:0.1:150:1800 \
   RRA:MIN:0.1:150:1800 \
   RRA:MAX:0.1:150:1800 \
   RRA:AVERAGE:0.1:450:1800 \
   RRA:MIN:0.1:450:1800 \
   RRA:MAX:0.1:150:1800

La création des fichiers est à faire une seule fois sur le serveur. Si la création est faite dans un répertoire /var/lib/collectd/rrd alors on peut facilement utiliser Collectd Graph Panel pour l'affichage des graphes.

Collecte des informations

Sur le serveur, une page fait l'extraction des paramètres de la requête POST et remplit la base de données RRDtool.

L'agent envoit les informations suivantes:

  • date: le temps Unix correspondant à la première mesure,
  • end: le temps Unix de la dernière mesure,
  • hc: la valeur du compteur sur les heures creuses,
  • hp: la valeur du compteur sur les heures pleines,
  • ic: le courant instantané,
  • pap: la puissance apparente.

Comme l'agent envoie les données par lot de 150 valeurs (ou plus si il y a eu des problèmes de connection), la mise à jour se fait en insérant plusieurs valeurs à la fois. Dans ce cas, rrdupdate s'attend à avoir le timestamp Unix suivit des valeurs des deux data sources (courant et puissance). Voici un extrait de la commande:

rrdupdate \
  /var/lib/collectd/rrd/home/teleinfo/teleinfo_power-home.rrd \
  1379885272:4:1040 1379885274:4:1040 1379885276:4:1040 \
  1379885278:4:1040 1379885280:4:1040 1379885282:4:1040 \
  1379885284:4:1040 1379885286:4:1040 1379885288:4:1040 ...

Pour l'installation de la collecte, copier le fichier edf-collect.php sur le serveur en s'arrangeant pour rendre accessible la page via le serveur web. Source: edf-collect.php.txt

Affichage des informations

Collectd Graph Panel est une application web écrite en PHP et Javascript permettant d'afficher les graphes collectés par collectd. Si les graphes sont créés au bon endroit, alors cette application les reconnaitra et permettra de les afficher. Pour cela, il faut ajouter le plugin teleinfo.php dans le répertoire plugin. Source: teleinfo.php.txt

unzip CGP-0.4.1.zip
cp teleinfo.php.txt CGP-0.4.1/plugin/teleinfo.php

Voici le résultat (nettement améliorable mais c'est un premier pas)...

Et maintenant

Voir sa consommation éléctrique a un petit coté ludique. Parfois c'est surprenant de constater que la consommation éléctrique ne descend pas en dessous de 200W. Ceci dit c'est normal avec toutes ces Box, décodeurs, switch et autres appareils qui même en veille consomme quelques watts.

Suiviconso

Planete Domotique

Integration of Ada Web Server behind an Apache Server

By Stephane Carrez 2013-07-07 07:54:25

When you run several web applications implemented in various languages (php, Java, Ada), you end up with some integration issue. The PHP application runs within an Apache Server, the Java application must runs in a Java web server (Tomcat, Jetty), and the Ada application executes within the Ada Web Server. Each of these web servers need a distinct listening port or distinct IP address. Integration of several web servers on the same host, is often done by using a front-end server that handles all incomming requests and dispatches them if necessary to other web servers.

In this article I describe the way I have integrated the Ada Web Server. The Apache Server is the front-end server that serves the PHP files as well as the static files and it redirects some requests to the Ada Web Server.

Virtual host definition

The Apache Server can run more than one web site on a single machine. The Virtual hosts can be IP-based or name-based. We will use the later because it provides a greater scalability. The virtual host definition is bound to the server IP address and the listening port.

<VirtualHost *:80>
  ServerAdmin webmaster@localhost
  ServerAlias demo.vacs.fr
  ServerName demo.vacs.fr
...
  LogLevel warn
  ErrorLog /var/log/apache2/demo-error.log
  CustomLog /var/log/apache2/demo-access.log combined
</VirtualHost>

The ServerName part is matched against the Host: request header that is received by the Apache server.

The ErrorLog and CustomLog are not part of the virtual hosts definition but they allow to use dedicated logs which is useful for trouble shotting issues.

Setting up the proxy

The Apache mod_proxy module must be enabled. This is the module that will redirect the incomming requests to the Ada Web Server.

  <Proxy *>
    AddDefaultCharset off
    Order allow,deny
    Allow from all
  </Proxy>

Redirection rules

The Apache mod_rewrite module must be enabled.

  RewriteEngine On

A first set of rewriting rules will redirect the request to dynamic pages to the Ada Web Server. The [P] flag activates the proxy and redirects the request. The Ada Web Server is running on the same host but is using port 8080.

  # Let AWS serve the dynamic HTML pages.
  RewriteRule ^/demo/(.*).html$ http://localhost:8080/demo/$1.html [P]
  RewriteRule ^/demo/auth/(.*)$ http://localhost:8080/demo/auth/$1 [P]
  RewriteRule ^/demo/statistics.xml$ http://localhost:8080/demo/statistics.xml [P]

When the request is redirected, the mod_proxy will add a set of headers that can be used within AWS if necessary.

Via: 1.1 demo.vacs.fr
X-Forwarded-For: 31.39.214.181
X-Forwarded-Host: demo.vacs.fr
X-Forwarded-Server: demo.vacs.fr

The X-Forwarded-For: header indicates the IP address of client.

Static files

Static files like images, CSS and javascript files can be served by the Apache front-end server. This is faster than proxying these requests to the Ada Web Server. At the same time we can setup some expiration and cache headers sent in the response (Expires: and Cache-Control: respectively). The definition below only deal with images that are accessed from the /demo/images/ URL component. The Alias directive tells you how to map the URL to the directory on the file system that holds the files.

  Alias /demo/images/ "/home/htdocs.demo/web/images/"
  <Directory "/home/htdocs.demo/web/images/">
    Options -Indexes +FollowSymLinks

    # Do not check for .htaccess (perf. improvement)
    AllowOverride None
    Order allow,deny
    allow from all
                                                 
    # enable expirations
    ExpiresActive On
                                  
    # Activate the browser caching
    # (CSS, images and scripts should not change)
    ExpiresByType image/png A1296000
    ExpiresByType image/gif A1296000
    ExpiresByType image/jpg A1296000
  </Directory>

This kind of definition is repeated for each set of static files (javascript and css).

Proxy Overhead

The proxy adds a small overhead that you can measure by using the Apache Benchmark tool. A first run is done on AWS and another on Apache.

ab -n 1000 http://localhost:8080/demo/compute.html
ab -n 1000 http://demo.vacs.fr/demo/compute.html

The overhead will depend on the application and the page being served. On this machine, the AWS server can process arround 720 requests/sec and this is reduced to 550 requests/sec through the Apache front-end (23% decrease).

Bacula database cleanup

By Stephane Carrez 2013-06-30 07:41:49

Bacula maintains a catalog of files in a database. Over time, the database grows and despite some automatic purge and job cleanup, some information remains that is no longer necessary. This article explains how to remove some dead records from the Bacula catalog.

Bacula maintains a list of backup jobs that have been executed in the job table. For each job, it keeps the list of files that have been saved in the file table. When you do a restore, you somehow select the job to restore and pick files from that job. There should not exist any file entry associated with a non existing job. Unfortunately this is not the case. I've found that some files (more than 2 millions entries) were pointing to some job that did not exist.

Discovering dead jobs still referenced

The first step is to find out which job has been deleted and is still referenced by the file table. First, let's create a temporary table that will hold the job ids associated with the files.

mysql> create temporary table job_files (id bigint);

The use of a temporary table was necessary in my case because the file table is so big and the ReadyNAS so slow that scanning the database takes too much time.

Now, we can populate the temporary table with the job ids:

mysql> insert into job_files select distinct file.jobid from file;
Query OK, 350 rows affected (8 min 53.26 sec)
Records: 350  Duplicates: 0  Warnings: 0

The list of jobs that have been removed but are still referenced by a file is obtained by:

mysql> select job_files.id from job_files
 left join job on job_files.id = job.jobid
 where job.jobid is null;
+------+
| id   |
+------+
| 2254 | 
| 2806 | 
+------+
2 rows in set (0.05 sec)

Deleting Dead Files

Deleting all the file records in one blow was not possible for me because there was too many files to delete and the mysql server did not have enough resources on the ReadyNAS to do it. I had to delete these records in batch of 100000 files, the process was repeated several times (each delete query took more than 2mn!!!).

mysql> delete from file where jobid = 2254 limit 100000;

Conclusion

This cleanup process allowed me to reduce the size of the file table from 10 millions entries to 7 millions. This improves the database performance and speeds up the Bacula catalog backup process.

Optimization with Valgrind Massif and Cachegrind

By Stephane Carrez 2013-03-02 16:19:53

Memory optimization reveals sometimes some nice surprise. I was interested to analyze the memory used by the Ada Server Faces framework. For this I've profiled the unit tests program. This includes 130 tests that cover almost all the features of the framework.

Memory analysis with Valgrind Massif

Massif is a Valgrind tool that is used for heap analysis. It does not require the application to be re-compiled and can be used easily. The application is executed by using Valgrind and its tool Massif. The command that I've used was:

valgrind --tool=massif --threshold=0.1 \
   --detailed-freq=1 --alloc-fn=__gnat_malloc \
   bin/asf_harness -config test.properties

The valgrind tool creates a file massif.out.NNN which contains the analysis. The massif-visualizeris a graphical tool that reads the file and allows you to analyze the results. It is launched as follows:

massif-visualizer massif.out.19813

(the number is the pid of the process that was running, replace it accordingly).

The tool provides a graphical representation of memory used over the time. It allows to highlight a given memory snapshot and understand roughly where the memory is used.

Memory consumption with Massif [before]

While looking at the result, I was intrigued by a 1MB allocation that was made several times and then released (It creates these visual spikes and it correspond to the big red horizontal bar that appears visually). It was within the sax-utils.adb file that is part of the XML/Ada library. Looking at the implementation, it turns out that it allocates a hash table with 65536 entries. This allocation is done each time the sax parser is created. I've reduced the size of this hash table to 1024 entries. If you want to do it, change the following line in sax/sax-symbols.ads (line 99):

   Hash_Num : constant := 2**16;

by:

   Hash_Num : constant := 2**10;

After building, checking there is no regression (yes, it works), I've re-run the Massif tool and here are the results.

Memory consumption with Massif [after]

The peak memory was reduced from 2.7Mb to 2.0Mb. The memory usage is now easier to understand and analyse because the 1Mb allocation is gone. Other memory allocations have more importance now. But wait. There is more! My program is now faster!

Cache analysis with cachegrind

To understand why the program is now faster, I've used Cachegrind that measures processor cache performance. Cachegrind is a cache and branch-prediction profiler provided by Valgrind as another tool. I've executed the tool with the following command:

valgrind --tool=cachegrind \
    bin/asf_harness -config test.properties

I've launched it once before the hash table correction and once after. Similar to Massif, Cachegrind generates a file cachgrind.NNN that contains the analysis. You analyze the result by using either cg_annotate or kcachegrind. Having two Cachegrind files, I've used cg_diffto somehow get diff between the two executions.

cg_diff cachegrind.out.24198 cachegrind.out.23286 > cg.out.1
cg_annotate cg.out.1

Before the fix, we can see in Cachegrind report that the most intensive memory operations are performed by Sax.Htable.Reset operation and by the GNAT operation that initializes the Sax.Symbols.Symbol_Table_Record type which contains the big hash table. Dr is the number of data reads, D1mr the L1 cache read miss and Dw is the number of writes with D1mw representing the L1 cache write miss. Having a lot of cache miss will slow down the execution: L1 cache access requires a few cycles while main memory access could cost several hundreds of them.

--------------------------------------------------------------------------------
         Dr      D1mr          Dw      D1mw 
--------------------------------------------------------------------------------
212,746,571 2,787,355 144,880,212 2,469,782  PROGRAM TOTALS

--------------------------------------------------------------------------------
        Dr      D1mr         Dw      D1mw  file:function
--------------------------------------------------------------------------------
25,000,929 2,081,943     27,672       244  sax/sax-htable.adb:sax__symbols__string_htable__reset
       508       127 33,293,050 2,080,768  sax/sax-htable.adb:sax__symbols__symbol_table_recordIP
43,894,931   129,786  7,532,775     8,677  ???:???
15,021,128     4,140  5,632,923         0  pthread_getspecific
 7,510,564     2,995  7,510,564    10,673  ???:system__task_primitives__operations__specific__selfXnn
 6,134,652    41,357  4,320,817    49,207  _int_malloc
 4,774,547    22,969  1,956,568     4,392  _int_free
 3,753,930         0  5,630,895     5,039  ???:system__task_primitives__operations(short,...)(long, float)

With a smaller hash table, the Cachegrind report indicates a reduction of 24,543,482 data reads and 32,765,323 data writes. The cache read miss was reduced by 2,086,579 (74%) and the cache write miss was also reduced by 2,056,247 (83% reduction!).

With a small hash table, the Sax.Symbols.Symbol_Table_Record gets initialized quicker and its cleaning needs less memory accesses, hence CPU cycles. By having a smaller hash table, we also benefit from less cache miss: using a 1Mb hash table flushes a big part of the data cache.

--------------------------------------------------------------------------------
         Dr    D1mr          Dw    D1mw 
--------------------------------------------------------------------------------
188,203,089 700,776 112,114,889 413,535  PROGRAM TOTALS

--------------------------------------------------------------------------------
        Dr    D1mr        Dw   D1mw  file:function
--------------------------------------------------------------------------------
43,904,760 120,883 7,532,577  8,407  ???:???
15,028,328      98 5,635,623      0  pthread_getspecific
 7,514,164     288 7,514,164  9,929  ???:system__task_primitives__operations__specific__selfXnn
 6,129,019  39,636 4,305,043 48,446  _int_malloc
 4,784,026  18,626 1,959,387  3,261  _int_free
 3,755,730       0 5,633,595  4,390  ???:system__task_primitives__operations(short,...)(long, float)
 2,418,778      65 2,705,140     14  ???:system__tasking__initialization__abort_undefer
 3,839,603   2,605 1,283,289      0  malloc

Conclusion

Running massif and cachegrind is very easy but it may take some time to figure out how to understand and use the results. A big hash table is not always a good thing for an application. By creating cache misses it may in fact slow down the application. To learn more about this subject, I recommend the excellent document What Every Programmer Should Know About Memory written by Ulrich Drepper.

Ada Web Application 0.3.0 is available

By Stephane Carrez 2013-02-15 20:45:46

Ada Web Application is a framework to build web applications.

  • AWA uses Ada Server Faces for the web framework. This framework is using several patterns from the Java world such as Java Server Faces and Java Servlets.
  • AWA provides a set of ready to use and extendable modules that are common to many web application. This includes managing the login, authentication, users, permissions.
  • AWA uses an Object Relational Mapping that helps in writing Ada applications on top of MySQL or SQLite databases. The ADO framework allows to map database objects into Ada records and access them easily.
  • AWA is a model driven engineering framework that allows to design the application data model using UML and generate the corresponding Ada code.

Ada Web Application Architecture

The new version of AWA provides:

  • New jobs plugin to manage asynchronous jobs,
  • New storage plugin to manage a storage space for application documents,
  • New votes plugin to allow voting on items,
  • New question plugin to provide a general purpose Q&A.

AWA can be downloaded at http://code.google.com/p/ada-awa/downloads/list

A live demonstration of various features provided by AWA is available at http://demo.vacs.fr/atlas

Dynamo 0.6.0 is available

By Stephane Carrez 2013-02-10 21:57:02

Dynamo is a tool to help developers write some types of Ada Applications which use the Ada Server Faces or Ada Database Objects frameworks. Dynamo provides several commands to perform one specific task in the development process: creation of an application, generation of database model, generation of Ada model, creation of database.

The new version of Dynamo provides:

  • A new command build-doc to extract some documentation from the sources,
  • The generation of MySQL and SQLite schemas from UML models,
  • The generation of Ada database mappings from UML models,
  • The generation of Ada beans from the UML models,
  • A new project template for command line tools using ADO,
  • A new distribution command to merge the resource bundles.

The most important feature is probably the Ada code generation from a UML class diagram. With this, you can design the data model of an application using ArgoUML and generate the Ada model files that will be used to access the database easily through the Ada Database Objects library. The tool will also generate the SQL database schema so that everything is concistent from your UML model, to the Ada implementation and the database tables.

The short tutorial below indicates how to design a UML model with ArgoUML, generate the Ada model files, the SQL files and create the MySQL database.

The Dynamo tool is available at http://code.google.com/p/ada-gen.

To build Dynamo, you will need:

Ada Database Objects 0.4.0 is available

By Stephane Carrez 2013-02-10 21:38:42

The Ada Database Objects is an Object Relational Mapping for the Ada05 programming language. It allows to map database objects into Ada records and access databases easily. Most of the concepts developped for ADO come from the Java Hibernate ORM. ADO supports MySQL and SQLite databases.

The new version brings:

  • Support to reload query definitions,
  • It optimizes session factory implementation,
  • It allows to customize the MySQL database connection by using MySQL SET

This version can be downloaded at http://code.google.com/p/ada-ado/downloads/list.

Ada Server Faces 0.5.0 is available

By Stephane Carrez 2013-02-05 22:29:27

Ada Server Faces is an Ada implementation of several Java standard web frameworks.

  • The Java Servlet (JSR 315) defines the basis for a Java application to be plugged in Web servers. It standardizes the way an HTTP request and HTTP response are represented. It defines the mechanisms by which the requests and responses are passed from the Web server to the application possibly through some additional filters.
  • The Java Unified Expression Language (JSR 245) is a small expression language intended to be used in Web pages. Through the expressions, functions and methods it creates the link between the Web page template and the application data identified as beans.
  • The Java Server Faces (JSR 314 and JSR 344) is a component driven framework which provides a powerful mechanism for Web applications. Web pages are represented by facelet views (XHTML files) that are modelized as components when a request comes in. A lifecycle mechanism drives the request through converters and validators triggering events that are received by the application. Navigation rules define what result view must be rendered and returned.

Ada Server Faces gives to Ada developers a strong web framework which is frequently used in Java Web applications. On their hand, Java developers could benefit from the high performance that Ada brings: apart from the language, they will use the same design patterns.

Ada Server Faces

The new version of Ada Server Faces is available and brings the following changes:

  • The Security packages was moved in a separate project: Ada Security,
  • New demo to show OAuth and Facebook API integration,
  • Integrated jQuery 1.8.3 and jQuery UI 1.9.2,
  • New converter to display file sizes,
  • Javascript support was added for click-to-edit behavior,
  • Add support for JSF session beans,
  • Add support for servlet error page customization,
  • Allow navigation rules to handle exceptions raised by Ada bean actions,
  • Support the JSF 2.2 conditional navigation,
  • New functions fn:escapeXml and fn:replace.

The new version can be downloaded on the Ada Server Faces project page. A live demo is available at http://demo.vacs.fr/demo.

Ada Utility Library 1.6.1 is available

By Stephane Carrez 2013-02-03 22:06:08

Ada Utility Library is a collection of utility packages for Ada 2005. This release is provided as workarround release to avoid the gcc 4.7 bug 53737. It allows to build Ada Utility Library with gcc 4.7.2.

Ada Security 1.0.0 is available

By Stephane Carrez 2013-01-20 17:54:13

Ada Security is a security framework which allows web applications to define and enforce security policies. The framework allows users to authenticate by using OpenID Authentication 2.0. Ada Security also defines a set of client methods for using the OAuth 2.0 protocol.

Ada Security Framework

  • A security policy manager defines and implements the set of security rules that specify how to protect the system or the resources.
  • A user is authenticated in the application. Authentication can be based on OpenID or another system.
  • A security context holds the contextual information that allows the security policy manager to verify that the user is allowed to access the protected resource according to the policy rules.

The Ada Security framework can be downloaded at Ada Security project page.

The framework is the core security framework used by Ada Server Faces and Ada Web Application to protect access to resources.

Ada Utility Library 1.6.0 is available

By Stephane Carrez 2012-12-26 21:16:23

Ada Utility Library is a collection of utility packages for Ada 2005. A new version is available which provides:

  • Support for HTTP clients (curl, AWS, ...)
  • Support for REST APIs using JSON
  • New operations To_JSON and From_JSON for easy object map serialization
  • Added a listeners to help implementing the observer/listener design patterns
  • Added support for wildcard mapping in serialization framework
  • New option -d <dir> for the unit test harness to change the working directory,
  • New example facebook.adb to show the REST support.

It has been compiled and ported on Linux, Windows and Netbsd (gcc 4.4, GNAT 2011, gcc 4.6.3). You can download this new version at http://code.google.com/p/ada-util/downloads/list.

Disabling overlay scrollbar fixes the Thunderbird scrollbar position issue on Ubuntu 12.04

By Stephane Carrez 2012-12-09 08:55:11

Since Ubuntu 12.04 upgrade, the Thunderbird scrollbar position were not visible any more. The scrollbar works but you have no visual feedback to know where you are in your long lists. Annoying!!! It turns out that this was a feature of the overlay scrollbar.

To restore the Thunderbird scrollbar, remove the following packages which are causing these troubles:

sudo apt-get remove overlay-scrollbar liboverlay-scrollbar-0.2-0 liboverlay-scrollbar3-0.2-0

Since I've disabled it, I realized many others have the issue. The How do I disable overlay scrollbars? Q&A gives other hints.

Ada BFD 1.0 is available

By Stephane Carrez 2012-11-11 09:29:47

Ada BFD is an Ada binding for the GNU Binutils BFD library.

It allows to read binary ELF, COFF files by using the GNU BFD. The Ada BFD library allows to:

  • list and scan the ELF sections of an executable or object file,
  • get the content of the ELF sections,
  • get access to the symbol table,
  • use the BFD disassembler

Version 1.0 of this Ada binding library is now available on Ada BFD. This new release bring the following changes:

  • Fixed installation of library
  • Added examples for BfdAda
  • Add support to use the disassembler
");