gecko-dev/tools/page-loader

# This Source Code Form is subject to the terms of the Mozilla Public
# License, v. 2.0. If a copy of the MPL was not distributed with this
# file, You can obtain one at http://mozilla.org/MPL/2.0/.

Rough notes on setting up this test app. jrgm@netscape.com 2001/08/05

1) this is intended to be run as a mod_perl application under an Apache web
   server. [It is possible to run it as a cgi-bin, but then you will be paying
   the cost of forking perl and re-compiling all the required modules on each
   page load].

2) it should be possible to run this under Apache on win32, but I expect that
   there are *nix-oriented assumptions that have crept in. (You would also need
   a replacement for Time::HiRes, probably by using Win32::API to directly
   call into the system to Windows 'GetLocalTime()'.)

3) You need to have a few "non-standard" Perl Modules installed. This script
   will tell you which ones are not installed (let me know if I have left some
   out of this test).

--8<--------------------------------------------------------------------
#!/usr/bin/perl 
my @modules = qw{
    LWP::UserAgent   SQL::Statement     Text::CSV_XS      DBD::CSV          
    DBI              Time::HiRes        CGI::Request      URI               
    MIME::Base64     HTML::Parser       HTML::Tagset      Digest::MD5      
    };
for (@modules) {
    printf "%20s", $_;
    eval "use $_;";
    if ($@) {
        print ", I don't have that.\n";
    } else {
        print ", version: ", eval "\$" . "$_" . "::VERSION", "\n";
    }
}
--8<--------------------------------------------------------------------

   For modules that are missing, you can find them at http://www.cpan.org/.
   Download the .tar.gz files you need, and then (for the most part) just 
   do 'perl Makefile.PL; make; make test; make install'.

   [Update: 28-Mar-2003] I recently installed Redhat 7.2, as server, which
   installed Apache 1.3.20 with mod_perl 1.24 and perl 5.6.0. I then ran the
   CPAN shell (`perl -MCPAN -e shell') and after completing configuration, I
   did 'install Bundle::CPAN', 'install Bundle::LWP' and 'install DBI' to
   upgrade tose modules and their dependencies. These instructions work on OSX
   as well, make sure you run the CPAN shell with sudo so you have sufficient
   privs to install the files.

   CGI::Request seems to have disappeared from CPAN, but you can get a copy
   from <http://stein.cshl.org/WWW/software/CGI::modules/> and then install
   with the standard `perl Makefile.PL; make; make test; make install'.

   To install the SQL::Statement, Text::CSV_XS, and DBD::CSV modules, there is
   a bundle available on CPAN, so you can use the CPAN shell and just enter
   'install Bundle::DBD::CSV'.

   At the end of this, the output for the test program above was the
   following.  (Note: you don't necessarily have to have the exact version
   numbers for these modules, as far as I know, but something close would be
   safest).

      LWP::UserAgent, version: 2.003
      SQL::Statement, version: 1.005
        Text::CSV_XS, version: 0.23
            DBD::CSV, version: 0.2002
                 DBI, version: 1.35
         Time::HiRes, version: 1.43
        CGI::Request, version: 2.75
                 URI, version: 1.23
        MIME::Base64, version: 2.18
        HTML::Parser, version: 3.27
        HTML::Tagset, version: 3.03
         Digest::MD5, version: 2.24

4) There is code to draw a sorted graph of the final results, but I have
   disabled the place in 'report.pl' where its use would be triggered (look
   for the comment). This is so that you can run this without having gone
   through the additional setup of the 'gd' library, and the modules GD and
   GD::Graph. If you have those in place, you can turn this on by just
   reenabling the print statement in report.pl

   [Note - 28-Mar-2003: with Redhat 7.2, libgd.so.1.8.4 is preinstalled to
   /usr/lib. The current GD.pm modules require libgd 2.0.5 or higher, but you
   use 1.8.4 if you install GD.pm version 1.40, which is available at
   <http://stein.cshl.org/WWW/software/GD/old/GD-1.40.tar.gz>. Just do 'perl
   Makefile.PL; make; make install' as usual. I chose to build with JPEG
   support, but without FreeType, XPM and GIF support. I had a test error when
   running 'make test', but it works fine for my purposes. I then installed
   'GD::Text' and 'GD::Graph' from the CPAN shell.]

5) To set this up with Apache, create a directory in the cgi-bin for the web
   server called e.g. 'page-loader'.

5a) For Apache 1.x/mod_perl 1.x, place this in the Apache httpd.conf file,
    and skip to step 5c.

--8<--------------------------------------------------------------------
Alias /page-loader/  /var/www/cgi-bin/page-loader/
<Location /page-loader>
SetHandler  perl-script
PerlHandler Apache::Registry
PerlSendHeader On
Options +ExecCGI
</Location>
--8<--------------------------------------------------------------------

    [MacOSX note: The CGI folder lives in /Library/WebServer/CGI-Executables/
    so the Alias line above should instead read:

      Alias /page-loader/  /Library/WebServer/CGI-Executables/page-loader
    
    Case is important (even though the file system is case-insensitive) and
    if you type it incorrectly you will get "Forbidden" HTTP errors.
    
    In addition, perl (and mod_perl) aren't enabled by default. You need to 
    uncomment two lines in httpd.conf:
      LoadModule perl_module        libexec/httpd/libperl.so
      AddModule mod_perl.c
    (basically just search for "perl" and uncomment the lines you find).]
  
5b) If you're using Apache 2.x and mod_perl 1.99/2.x (tested with Red Hat 9),
    place this in your perl.conf or httpd.conf:

--8<--------------------------------------------------------------------
Alias /page-loader/  /var/www/cgi-bin/page-loader/

<Location /page-loader>
SetHandler perl-script
PerlResponseHandler ModPerl::RegistryPrefork
PerlOptions +ParseHeaders
Options +ExecCGI
</Location>
--8<--------------------------------------------------------------------

   If your mod_perl version is less than 1.99_09, then copy RegistryPrefork.pm
   to your vendor_perl ModPerl directory (for example, on Red Hat 9, this is
   /usr/lib/perl5/vendor_perl/5.8.0/i386-linux-thread-multi/ModPerl).

   If you are using mod_perl 1.99_09 or above, grab RegistryPrefork.pm from
   http://perl.apache.org/docs/2.0/user/porting/compat.html#C_Apache__Registry___C_Apache__PerlRun__and_Friends
   and copy it to the vendor_perl directory as described above.

5c) When you're finished, restart Apache.  Now you can run this as
    'http://yourserver.domain.com/page-loader/loader.pl'
    
6) You need to create a subdirectory call 'db' under the 'page-loader'
   directory. This subdirectory 'db' must be writeable by UID that Apache
   executes as (e.g., 'nobody' or 'apache'). [You may want to figure out some
   other way to do this if this web server is not behind a firewall].

7) You need to assemble a set of content pages, with all images, included JS
   and CSS pulled to the same directory. These pages can live anywhere on the
   same HTTP server that is running this app. The app assumes that each page
   is in its own sub-directory, with included content below that
   directory. You can set the location and the list of pages in the file
   'urllist.txt'. [See 'urllist.txt' for further details on what needs to be
   set there.]

   There are various tools that will pull in complete copies of web pages
   (e.g. 'wget' or something handrolled from LWP::UserAgent). You should edit
   the pages to remove any redirects, popup windows, and possibly any platform
   specific JS rules (e.g., Mac specific CSS included with
   'document.write("LINK...'). You should also check that for missing content,
   or URLs that did not get changed to point to the local content. [One way to
   check for this is tweak this simple proxy server to check your links:
   http://www.stonehenge.com/merlyn/WebTechniques/col34.listing.txt)
   
   [MacOSX note: The web files live in /Library/WebServer/Documents, so you will
   need to modify urllist.txt to have the appropriate FILEBASE and HTTPBASE.]

8) The "hook" into the content is a single line in each top-level document like this:
      <!-- MOZ_INSERT_CONTENT_HOOK -->
   which should be placed immediately after the opening <HEAD> element. The script uses
   this as the way to substitute a BASE HREF and some JS into the page which will control
   the exectution of the test.

9) You will most likely need to remove all load event handlers from your
   test documents (onload attribute on body and handlers added with
   addEventListener).

10) Because the system uses (X)HTML base, and some XML constructs are not
    subject to that (for example xml-stylesheet processing instructions),
    you may need to provide the absolute path to external resources.

11) If your documents are tranformed on the client side with XSLT, you will
    need to add this snippet of XSLT to your stylesheet (and possibly make
    sure it does not conflict with your other rules):
--8<--------------------------------------------------------------------
<!-- Page Loader -->
<xsl:template match="html:script">
  <xsl:copy>
  <xsl:apply-templates/>
  </xsl:copy>
  <xsl:for-each select="@*">
    <xsl:copy/>
  </xsl:for-each>
</xsl:template>
--8<--------------------------------------------------------------------
    And near the top of your output rules add:
       <xsl:apply-templates select="html:script"/>
    Finally make sure you define the XHTML namespace in the stylesheet
    with "html" prefix.

12) I've probably left some stuff out. Bug jrgm@netscape.com for the missing stuff.