creacog

Product Owner/Manager; Digital Project Manager, ex-Developer Available for Product Owner roles

Jul

2011

Which width? (AIR on devices)

Within the Flash APIs is there there are many similarly named properties which provide subtly different results based on their spec and the state of the application. Width and height are such properties.

Having just read of Mike Jones’ recent article: 10 Tips When Developing For Multiple Devices, and tip #1: Check screen dimension on initialisation, it includes a couple of lines of code to get the landscape height and width of the device screen using a boolean expression to determine the longest dimension. For me this begs the question, why can’t I simply get the right values directly from the API? Especially if my app.xml properties state the app should only display landscape, full screen.

The properties examined here:

Although accurately described in the reference, it isn’t necessarily easy to visualise the differences in behaviour for these properties from the documentation alone. So herewith, a simple HelloWidthHeight app to trace out the values upon initialisation. Included in that app is a class implementing logic based on Mike’s as static methods. The results of debug sessions executed on my HTC Desire are detailed for comparison in the table below.

Key app.xml <initialWindow> properties used were:

<visible>true</visible>
<aspectRatio>landscape</aspectRatio>
<autoOrients>false</autoOrients>
<fullScreen>true</fullScreen>

	A	B	C	D
stage.orientation	default	rotatedRight	rotatedRight	rotatedRight
stage.width	0	0	0	0
stage.height	0	0	0	0
stage.stageWidth	480	800	800	800
stage.stageHeight	800	480	480	480
stage.fullScreenWidth	480	480	800	480
stage.fullScreenHeight	800	800	480	800
Oriented.landscapeScreenWidth	800	800	800	800
Oriented.landscapeScreenHeight	480	480	480	480
Oriented.portraitScreenWidth	480	480	480	480
Oriented.portraitScreenHeight	800	800	800	800

Screen inactive, phone held in any orientation
Screen active, phone lying flat on the desk
Screen active, phone held landscape, mic to the right (this is the intended orientation for the app)
Screen active, phone held portrait, mic to the bottom

In conclusion, unless the application is only ever launched while the screen is active (though that would be the norm), the safe way of determining an oriented width and height is to use a boolean expression along with the fullScreenWidth and fullScreenHeight properties.

UPDATE (20-July-2011): The following table runs the same code in the BlackBerry PlayBook Simulator:

	A	B
stage.orientation	default	default
stage.width	0	0
stage.height	0	0
stage.stageWidth	1024	1024
stage.stageHeight	600	600
stage.fullScreenWidth	1024	1024
stage.fullScreenHeight	600	600
Oriented.landscapeScreenWidth	1024	1024
Oriented.landscapeScreenHeight	600	600
Oriented.portraitScreenWidth	600	600
Oriented.portraitScreenHeight	1024	1024

Simulator initially landscape
Simulator initially portrait

Note: Behaviour is slightly different on the PlayBook simulator in that the orientation of the application is set before the initialisation code runs. Therefore unlike on the HTC Desire, fullScreenWidth and fullScreenHeight are correctly returned as the orientation of the application expects.

If anyone else cares to run the code, I’d be interested to hear how variable other devices are in their results, or if this is showing a bug when run on HTC Desire. Running AIR 2.7 here.

Feb

2011

Apple Cinema HD 23inch dies, then resurrects

This morning I was greeted with a blank screen and the power light flashing short-long-short, and groaned at how expensive today was about to get.

The Apple support page states that this flash pattern indicates the wrong power adaptor is being used with the display. I’m using the original 90W adaptor supplied with the screen and keep it powered via an APC UPS.

Fortunately with a bit of googling I found the ‘paper w’ solution here. Essentially by using a strip of paper to blank off the middle of the 5 output pins from the power brick, the display is back to life.

From reading through comments made by others in both forums, and scanning over the power adaptors page, I’m forming the following conclusions. (Note: these are my guesses based only on my interpretation of other’s anecdotal evidence rather than anything authoritative).

1. I suspect the middle pin is used only to allow the monitor to sense which of the 3 Apple power-bricks is attached. If so then it seems more concerned with protecting the monitor from a low power brick, rather than from being overloaded somehow.

2. A number of people have indicated that simply replacing the brick like for like or with the higher power version makes no difference. This suggests to me that the fault is within the electronics of the monitor, when detecting the power source at switch-on, rather than in the brick.

So while this is far from an ideal solution, it is cheap, simple and quick to implement compared with trucking everything back for repair. However if in doubt, get it repaired by Apple – don’t blame me if your house burns down.

Feb

2011

Finally using a decent mouse-surface

I’ve been using Apple Magic Mouse for some time. I love it. I’ve used loads of optical mice before it. The one thing that annoys me, even with this mouse, is that it’s accuracy depends very much on finding a decent surface.

I tried loads and loads of surfaces. More or less any surface (except transparent) will do for short stints. But for my workstation, my most used have been the cover of a hard back note-book (such as the Black n’ Red Casebound Hardback A4 Book), glossy magazine covers, other cardboard rectangles (e.g. the back of a note pad) taped to the desk. All of these surfaces tend to suffer the following problems:

They wear out
They move across the desk unless taped down
They are abrasive t the mouse – the tracks on my magic mouse are looking a little worn
They get grubby
They they turn to mush if coffee is spilled on them
Eventually the only relationship of the screen pointer to the mouse is that it moved

Now I think I’ve finally found a surface that will not wear down my mouse rails any further, so far yields extremely accurate results and will stand up to a tea spillage. Unfortunately costs a few quid, but that’s better than trying to accommodate a pointer with a mind of it’s own when on a deadline. The Razer Vespula Gaming Mat.

Feb

2011

cPanel and mail subject filtering

If you are using ‘User level filtering’ of email in cPanel (v11.28.83), applied to the Subject header of an email, you may encounter circumstances where sometimes the filter fails for no obvious reason. On investigation of the email headers, it turned out that a certain sender was folding email subject headers over multiple lines. If the phrase being searched for was split over two lines within the header, the filter would fail.

For example, with rule : Subject contains ‘testing 123’

1. this would work:
Subject: line 1 testing 123

2. this would work:
Subject: line 1 line 2 testing 123

3. this would NOT work:
Subject: line 1 testing 123 line 2

It turns out that this is because cPanel simply passes the filter information down to exim, and it is here that the behaviour manifests.

I have posted an entry to the feature requests forum of cPanel, for it to be configured to better tolerate multiline subject headers. If you are encountering this problem, it would be helpful if you would contribute to that thread.

In the meantime a workaround is to use the ‘matches regex’ filter, replacing spaces in your match-phrase with ‘\\\\s+’. For example the following rule works in each of the three samples above.

Rule: Subject matches regex testing\\\\s+123

Jul

2010

switched to svn via https on mac os x

I finally decided it is time to switch to accessing subversion repositories via apache and the https scheme rather than ‘file://’ scheme. Search revealed many articles touching the subject, but none providing exactly what I need. So documenting:

The problem and motivation
Up to date?
Getting apache to serve https
Configuring apache to serve subversion
Migrating existing working copies to the new scheme
Reference links to articles that helped

Problem and motivation

My main development machine is a Mac Pro running 10.6.4. My development projects are all kept outside my home directory on a second disc, one project per folder. Each project folder contains a subversion repository folder named ‘svn’ along with one or more working copies.

For the last few years I have checked out working copies using the ‘file://’ protocol. This seemed the simplest and most efficient approach in this single-user environment. I use svn clients such as SvnX, and Subclipse and occasionally the command line (for which having the svnbook to hand is a must). I have groaned previously about DreamWeaver’s subversion integration ‘attempt’ failing to support the file protocal (remains unchanged in CS5).

There are a couple of things I want to solve:

Firstly, I occasionally use VisualStudio in Windows7 running in parallels on this machine to work with ASP.NET. In the past I’ve simply pointed it at the working copy via the network drive. Of course this means that I can’t undertake svn operations through Win7 since that working copy’s URL is alien to the Windows 7 instance. Not too much of an issue since it’s only a matter of switching back to a Mac window to do svn business. However there is an annoying problem with VisualStudio’s code completion where it is unable to correlate markup within an .aspx file with it’s .aspx.cs file when the site is on a parallels network drive. To solve this I need to check out a working copy to the Windows7 local disk.

Second, I have taken to getting out and about with my MacBook Pro. If I want to work on the move I need to check-out a working copy.

For both these situations using the ‘file’ protocol is inappropriate. Attempts to check work back in are bound to generate svn errors as multiple svn systems attempt to obtain exclusive locks on the repository.

If however we can get to a single user/process touching repository files we can solve this problem – enter apache.

Up to date?

First thing I did was get my subversion installation up to date. At time of writing 1.6.12. Installer available from CollabNet. Just run the installer and follow the instructions. Note: this version installs into a different location to that installed by Apple. You may need to tell your client tools the location of the subversion to use.

That said, this isn’t going to help with DreamWeaver all that much. DW’s Subversion functionality is tightly tied to specific versions. Should you dare touch a working copy with a later version (which changes some of the meta data) DreamWeaver will cease to work with that working copy. More information at this Adobe technote. So for now as far as I am concerned, until Adobe start releasing ‘upgrader extensions‘, DW’s subversion functionality remains useless and turned off with ‘.svn’ files cloaked and svnX used for commits.

Getting apache to serve https

In the short term I have no plans to open access to my repositories via the net. However it is a future possibility so I think it worth getting going with https from the outset is worth it.

Mac OS X of course uses Apache for web-sharing. However it’s default state is not configured to serve SSL. To do so, we need a secure certificate and some configuration changes. As I am the only person accessing this machine, and I trust myself, I have no need to obtain a certificate from a commercial authority.

Steps taken to create the certificate and configure apache to use it…

Create a certificate authority

mkdir /Library/Certs cd /Library/Certs perl /System/Library/OpenSSL/misc/CA.pl -newca [ENTER](to create new certificate)

Generate private key

openssl genrsa -des3 -out webserver.key 1024

generate a non-password protected copy of the key

openssl rsa -in webserver.key -out webserver.nopass.key

Generate a certificate request

openssl req -config /System/Library/OpenSSL/openssl.cnf \ -new -key webserver.key -out newreq.pem -days 3650

Sign the certificate request

perl /System/Library/OpenSSL/misc/CA.pl -signreq

You should now have created…

/Library/Certs/demoCA/ /Library/Certs/newcert.pem /Library/Certs/newreq.pem /Library/Certs/webserver.key /Library/Certs/webserver.nopass.key

Tell Apache to include SSL

We now need to edit apache’s httpd.conf. You need to ‘sudo’ to acquire sufficient privileges to do so, and need to take care. Optionally make a backup copy of httpd.conf.

cd /private/etc/apache2/ sudo cp httpd.conf httpd.conf.bak sudo pico httpd.conf

Find the following line and uncomment it by removing it’s # prefix

Include /private/etc/apache2/extra/httpd-ssl.conf

Use CTRL-O then CTRL-X to exit pico and we now need to edit the file we just included…

cd extra sudo pico httpd-ssl.conf

Go through the file finding the following attributes ensuring they are uncommented and point to the SSL files we just created…

SSLCertificateFile "/Library/Certs/newcert.pem" SSLCertificateKeyFile "/Library/Certs/webserver.nopass.key" SSLCACertificateFile "/Library/Certs/demoCA/cacert.pem" SSLCARevocationPath "/Library/Certs/demoCA/crl"

You should now be able to restart apache either through System Preferences… > Sharing > Web sharing, or

sudo apachectl graceful

You can access any errors via the Console application. If all is well you should be able to enter https://localhost/ into your browser’s address bar and get a result.

Configuring apache to connect to and serve subversion

Firstly, we are going to require a login. In my case I am going to create 3 login IDs. One for my normal workstation. Others for access via Win7 and my MacBook pro. So although all the work in the repro is by me, I can see which environment was used.
To do this, we will create an authorisation file containing the three users and place it somewhere sensible. For me…

cd /Volumes/projectdisc/projects mkdir subversion cd subversion mkdir authfile cd authfile sudo htpasswd -c svn_passwd mpuser sudo htpasswd svn_passwd mbpuser sudo htpasswd svn_passwd win7user

you should now have svn_passwd contianing 3 users and their password hash strings. Returning to apache configuration, you may have noticed that the last line of httpd.conf reads…

Include /private/etc/apache2/other/*.conf

This includes, in alphabetical order, any further files ending .conf in the subfolder ‘other’. We will use this to add a svn.conf file to that folder…

cd /private/etc/apache2/other sudo pico svn.conf

Will open an editor with empty/new file svn.conf. The first line of which will be:

LoadModule dav_svn_module /usr/libexec/apache2/mod_dav_svn.so

After this, we will add configuration blocks, one per project repository…

<Location /svn0000-svntest> DAV svn SVNPath /Volumes/projectdisc/projects/0000-svntest/svn AuthType Basic AuthName "subversion" AuthUserFile /Volumes/projectdisc/projects/subversion/authfile/svn_passwd Require valid-user SSLRequireSSL </Location>

Using keys ctrl-o then ctrl-x will save the new file and exit pico.

This configuration block tells apache to redirect svn0000-svntest to the repository at path /Volumes/projectdisc/projects/0000-svntest/svn. It requires a valid user authenticated against the file at /Volumes/projectdisc/projects/subversion/authfile/svn_passwd. Since this location will only ever be served through SSL, basic AuthType is ok and secure.

For apache to pick up this change, we need to restart it:

sudo apachectl graceful

One further step we need to make is to ensure apache is the only user/process to have control of the repository files. This shouldn’t be an issue as we are migrating access to be always via https and don’t want the file protocol used any more…

cd Volumes/projectdisc/projects/0000-svntest/ sudo chown -R www:www svn

At this point you should be able to access the repository through your browser with url http://localhost/svn0000-svntest

You will receive an alert indicating that the certificate is not trusted. You can tell safari that it should always trust this certificate.

To add further repositories, we simply need to change ownership as above, then add the detail to the configuration by simply duplicating the code block above with the Location text and SVNPath modified accordingly. The rest can remain as is.

Migrating existing working copies to the new scheme

My existing working copies each use the file:// url scheme. We need to convert them to use the new https scheme. This is pretty easy assuming you already have the Location added to svn.conf. Just cd to the working copy folder, use svn info to reveal the working copy’s url and relocate…

cd Volumes/projectdisc/projects/0000-svntest/wc svn info svn switch --relocate file:///Volumes/projectdisc/projects/0000-svntest/svn/trunk/wc https://localhost/svn0000-svntest/trunk/wc

If you are prompted that the certificate is invalid, use option p to permanently trust the certificate.

Reference links to articles that helped

Reading the following, articles that helped me work out what I wanted to do…

Along with chapters 3 and 10 of Subversion Version Control: Using The Subversion Version Control System in Development Projects ISBN-10: 0-13-185518-2 also on Safari books online.

Jun

2010

Adobe AIR 2 runtime s for Windows, Mac and Linux publicly available

Spreading the word based on Adobe’s notes…

The AIR 2 release includes many new features including:

Native Process API
Open documents with the user’s default application
Microphone data access
Mass storage device detection
Updated, faster WebKit with enhanced support for HTML5 and CSS3
Multi-touch
New networking support including UDP and server sockets
Screen reader support
Reduced CPU usage on idle
Up to 30% reduction in memory usage without recompiling an application
+ more

Links:

This coincides with the final release of Flash Player 10.1…

Jun

2010

hello desire

So, it’s good-bye to my trusty and slightly crumbling Nokia 3100 (recently a source of amusement and pity amongst my peers)…

Hello HTC Desire…

It could so easily have been hello iPhone. I do have iPod Touch which I enjoy using. But fundamentally as a Flash/ActionScript/Flex/AIR developer it made no sense at all to get a smartphone on which Flash has been nobbled.

(My own brief comment and observation on the iPhone/Flash debacle : It looks to me that both companies have incompatible business strategies with regard to delivery of RIAs on mobile devices. Discussion outside of the these strategies is in my opinion a deflection. It was disappointing to read Steve Job’s thoughts on Flash, which to my mind are ill-informed and based on half truths – out of character in those regards. It was also disappointing to watch Adobe CEO Shantanu Narayen’s response in interview with the Wall Street Journal where, in my humble opinion, he was far from convincing and by the end was sounding more like a parroting politician. Disappointing too that past quality and performance issues with Flash player gave Apple an easy ammunition to exaggerate and exploit. As an avid Apple Mac and Adobe Creative Suite user I hope the two companies can return to a professional relationship which doesn’t leave customers of both companies, like myself, out in the cold.)

So back to the HTC Desire, some first impressions…

The good

It looks good, feels good
Nice bright responsive screen
Call quality is good
Better quality camera than I expected
The main reason for getting this device – Flash based apps are allowed!

The not so good (compared with my iPod Touch experience)

There are too many buttons. I find myself pressing the wrong one most of the time. Sometimes a button press is required. Sometimes not. The whole thing is less intuitive than the iPod Touch with it’s single button.
It seems all too easy to initiate a call at random while scrolling through the contact list.
Text selection/cursor positioning is awful
There is no out of the box easy way of syncing Address book, Calendar, tunes, photos etc with my Macs. Looks like I need to purchase Missing Sync. That said, I did previously purchase Mobile Me to keep my Mac / Mac Book Pro and iPod Touch all in sync.
The Mail application is crap. I use a self-signed SSL certificate on my mail server, so I immediately hit the problem of a silent fail when trying to add connection details to the mail application. The hack in the forum thread worked in fixing it, i.e. turning off my router’s WAN connection, while inputting the connection details. Also it doesn’t list the mail folders on the server – all I get is the inbox. Apple’s Mail app by comparison is a doddle and reflects the structure of my mail account.

Fingers crossed for Android 2.2.

Anyway, looking forward to setting up some kind of tether to share the data connection with my MBP and more importantly getting something running in AIR for Android on there.

May

2010

Persits AspUpload, Content-type and Safari

A quick note for anyone using legacy server-side classic ASP and the Persits AspUpload component‘s sendBinary to download files to Apple’s Safari browser.

I ran into a problem recently where by Safari was appending “.html” to the file name of all downloads sent via the component. e.g. test.pdf would download as test.pdf.html. There are a couple of mentions on the apple support forums:

PDF files downloaded directly did not suffer this problem – suggesting that the server’s mime types seem ok. Other browsers did not suffer this problem. However the problem has to be server-side. i.e. the download via the script is being sent with Content-type text/html instead of application/pdf no matter what arguments I add to the sendBinary call.

The solution seems to be to manually set the script’s Response.contentType value. i.e. the following test script works…

<%@LANGUAGE="VBSCRIPT" CODEPAGE="65001"%>
<%
    FILEPATH = "D:\WWWRoot\sitepath\www\pdftest\TEST.pdf"
    Dim upl
    Set upl = Server.CreateObject("Persits.Upload.1")

    Response.ContentType = "application/pdf"

    upl.sendbinary FILEPATH, True, "application/pdf", True
%>

In my case this is using AspUpload version 3.0.0.2 on a Hostway Windows 2003 Gold plan, shared hosting.

Apr

2010

Flex 4 Spark VideoDisplay (measure, overlay, skin)

In a previous article I described some of the steps required in Flex 3 in order to keep an overlay position in sync with a VideoDisplay component. With the release of Flex 4, VideoDisplay is one of the components to be updated. Via VideoPlayer, it gets a Spark makeover. Internally its core, or ‘view’, has been reimplemented using Open Source Media Framework. For anyone wishing to keep an overlay in position over the view, however, the same key issue remains:

None of the various width or height properties of the VideoDisplay component describe the width or height of the contained view.

So, to get the properties of the view we need to delve deeper into VideoDisplay and use the mx_internal namespace. Even then, the ResizeEvent.RESIZE event is broadcast by VideoDisplay before it’s view dimensions (column B) have been updated to the new site.

In that previous article I observed that if, after each resize, I added an Event.EXIT_FRAME listener, I could guarantee that the VideoDisplay’s view had been sized by the time it was called. I observed that the same is true in Flex 4. However EXIT_FRAME didn’t feel like the right place to be doing this. So with a closer look at the component life cycle we can see that handling the FlexEvent.UPDATE_COMPLETE event is more appropriate. Column C shows the view’s width/height after UPDATE_COMPLETE has been broadcast. It colours red when a value disagrees with column B (RESIZE). And so, it is column C that contains the values we want to use while managing our overlay.

Using this information and Flex 4’s new component and skinning architecture, we can build quick and simple to use video overlay component rather than messing about with external listeners and calculating positions as we did in the previous Flex 3 examples.

Reviewing what Flex 4 includes ‘out of the box’…

VideoDisplay : a non-skinable component that wraps an osmf ‘view’. It translates much of the osmf api and events to what we are accustomed to with Flex 3.
VideoPlayer : A skinable component provides hooks for skin parts. A required part is an instance of VideoDisplay. The others are the buttons and controls we’d normally expect when giving the user control of the video.

The VideoPlayer and it’s skin are overkill for this example, so we’ll create something far simpler…

VideoDisplayOverlayContainer : A container component who’s children will overlay a VideoDisplay
VideoDisplayOverlayContainerSkin : A simple skin, providing the required videoDisplay skin part.

This way, overlaying the video is as simple as giving the container some child components and a layout (defaults to basic layout). In the example that follows we overlay a label instance to each of the four corners and the centre of the view…


<components:VideoDisplayOverlayContainer
    id     ="videoDisplayContainer"
    source ="VideoSampleForFlex.mp4"
    width  ="100%"
    height ="100%" 	>
    <s:Label text="A" top="2" left="2" backgroundColor="0xffffff" backgroundAlpha="0.4" />
    <s:Label text="B" top="2" right="2" backgroundColor="0xffffff" backgroundAlpha="0.4"  />
    <s:Label text="C" bottom="2" left="2" backgroundColor="0xffffff" backgroundAlpha="0.4"  />
    <s:Label text="D" bottom="2" right="2" backgroundColor="0xffffff" backgroundAlpha="0.4"  />
    <s:Label text="O" verticalCenter="0" horizontalCenter="0" backgroundColor="0xffffff" backgroundAlpha="0.4"  />
</components:VideoDisplayOverlayContainer>

View source

Mar

2010

Flex builder 4 help system (CHC) and backups

(Edit: Of course this should have been titled Flash Builder 4)

One of the changes with Flash builder over previous Flex Builder products is in the help system. The help files are no-longer bundled and indexed by eclipse help. Instead a new help application is installed which attempts to combine the documentation with community content.

I must admit to not receiving this application particularly well in the first instance. However I have seen it develop over a number of weeks and as an AIR application it is expected to continue to develop independently of major product release cycles.

Since it is a new product, some rough edges can be expected. There is already a technote (cpsid_83103) relating to getting context-sensitive help working correctly in a fresh install of Flash Builder 4.

Randy Nielsen’s post on using the help client provides a good orientation for new users.

However I think there are some other things that system administrators at least (and users who backup their user-directories – particularly to the cloud) should be aware of…

1. The many names of adobe help

On a Mac it installs application “Adobe Help” into /Applications/Adobe. The main application window when running is titled “Adobe Community Help”. And you will see this application widely referred to in Adobe communications as CHC or Adobe Community Help Client. If you ever look for the preferences files for the application, you’ll find them in ~/Library/preferences/chc.xxxx.1 (where xxxx is a 40 digit hex number). Note: missing is the usual “com.adobe.” prefix.

2. Downloading the help files.

The help files need to be downloaded to your local system in order for the application to work. The current version of the application (3.0.0.400) only displays local content if the network is not available. Otherwise it always displays content from the web. This seems slightly wasteful and slow. Though we can expect an option to default to local content irrespective of network connectivity in future versions.

3. Help file size

You can manually control the downloads via Preferences>Updater settings then “Manage now”. This will show a table of help files, whether they are current or not and their size. The size however is slightly misleading. It refers to the zip file size and therefore the network bandwidth required to download the help package. Once unzipped however, the disk-space used is a considerably larger number. I find that the help packages for Fb4 and associated stuff expand to use 563Mb.

4. Where the help files are stored

On a mac the help files are expanded and stored in the user’s preferences folder. ~/Library/Preferences/chc.xxxx.1/Local Store/Help

To me this is not the right place to be storing such content. Help content is not preferences and is not necessary ‘per user’. Preferences are usually backed up. There is absolutely no need to be backing up this content – to do so is a waste of storage, network bandwidth and machine time. Once all Adobe products are integrated with CHC, the help folder will be considerably larger than my current 563Mb.

In summary

Adobe Help, Adobe Community Help Client (CHC), is currently in its infancy as an application, but will be soon in the hands of many, many users. As an AIR application, the development team have far more flexibility to make releases between major product cycles. I hope and believe that the points I highlight (and location of help files in particular) will be addressed in coming releases – they have all been submitted to the team – not just blogged here. However they are blogged here because I think storage suppliers could be in for a field day unless sys-admins are made aware, and add rules to their backup process to omit the help files.

creacog

Which width? (AIR on devices)

Apple Cinema HD 23inch dies, then resurrects

Finally using a decent mouse-surface

cPanel and mail subject filtering