I was just creating a layout using an IRF9Z34NS with the D2PAK footprint. The TO-263-2_TabPin1 is incorrect.
To correctly align with the PMOS schematic symbol it should have the pin numbering: D:1, G:2, S:3
I have created my own copy with the corrections.
But…wait a minute.
But the drain IS the tab pin!
I think the drain = tab should normally be 2 or 3 or 4 or have no number, depending on how you look at it (which seems to be inconsistent across the industry. For example with some packages I have seen a cutoff & useless center pin labeled as pin 2, connected to the drain=tab labeled as pin 4.)
Of course we could just number the pins #7, #19, and #304. 
The industry manages to use every permutation of pin numbering for 3 pin transistors and mosfets
The fact that TO-263-2 has the shortest name probably indicates that this is the most common configuration. You should be using the numbering in the datasheet and using a symbol that suits that numbering. There are all permutations of numbering in the symbol library.
I just went to the datasheet:
What I see there backs up the argument that all of this stuff is very inconsistent:
Note that there is a pin 2 which most of us would not try to use.
In my opinion, the only good way to deal with this situation is to choose your preferred pin number assignments for the footprint. Then look at the datasheet and make sure that your symbol agrees with the physical location of the pin on the footprint.
For a MOSFET in Dpak or D2Pak with the most common pinout, I would label the pins as I show in red:
And this is my TO263 footprint which I guess I really ought to call a 3 pin:
Bobs_TO263_2_Pin_Fat_01.kicad_mod (1.6 KB)
And herein lies the problem, it is very inconsistent.
Maybe the schematic symbol should change its pin designations, or at least identify the numbering on the schematic - currently it only shows GDS, this might help to avoid this confusion.
Maybe there should be a note or warning attached to all 3 pin devices to check the footprint against the datasheet before committing to copper. This has caught me out before in the past with THT transistors.
In my opinion that is reasonable. Even the SOT23 has numbering variations on some datasheets. But most of us who have been doing this for a while are aware and know to be careful.
You mean like this?
where the numbering of the pins is shown and also the name of the symbol has the particular permutation. There are symbols for the other 5 permutations, and of course for the N devices.
The footprint may be unsuitable for your symbol and device but it is not per se incorrect. You just have to choose the right footprint. If there isn’t one suitable, then modify an existing one or make one.
Or to make it a more general rule, check everything yourself. Pin that to your workshop wall.
In my opinion the numbering used by Infineon (GDS = 123) is more logical. Here’s that datasheet again.
If you use the 1234 numbering of International Rectifier, you end up with a useless pin 2 because it can’t be soldered to its pad even though it’s internally connected to pin 4. So you end up having to pick a footprint that has pads 134. Or use a footprint with pads 1234, then route a useless track to pad 2, or suppress the pad 2 not connected error in DRC.
Bob’s proposed numbering of pins would also work, but you have to pick the GSD symbol for your schematic.
Well, don’t say you weren’t forewarned, you admit to having similar experiences with THT transistors. Most of my THT transistors are EBC, but there are ECB ones, like the 2SC815, which caught me out once. Fortunately with THT transistors, you can twist the leads. Won’t be possible with your D-PAK P-MOSFET.
On this forum there was a long topic with the notorious 2N2222 in SMD form where the manufacturer had used a different numbering and the user was spitting chips (or transistors?) when the board didn’t work.
Did I already mention always check everything yourself? Or maybe “Check it us”, as an English translation in a Madrid optometrist read. 
In the symbol list the, I guess generic, PMOS is represented like this, the pin numbers are not shown and the only way to find them is to open the symbol in the editor. Sure I could look for another symbol but with a gazillion different mosfets I would probably end up having to create a symbol, hence the generic symbol with my particular device.
Having the pin numbers as well as the pin function would hopefully help to avoid footprint confusion.
Point noted about the useless pin 2.
Works for me. I placed the symbol Q_PMOS_GSD from the library Device on my schematic and got this:
I can also select a pin and see its properties in the status bar at the bottom.
So I don’t know why you are not seeing pin numbers in your schematic. My BJTs also show pin numbers.
I always work in reverse for these types of components.
I find the suitable footprint:
Match that footprint to the Data sheet and get pin 1 = G, 2 = D, 3 = S.
Next I go looking in the symbols for the PMOS with Pins GDS which when placed on the schematic shows as 1 = G, 2 = D & 3 = S.
In case some have never noticed, the order of the letters in the symbols, eg. Q_PJFET_ DSG or Q_NPN_BCE is always shown in pin order 1, 2 & 3.
Ok, I spotted why - I picked what I thought was a generic PMOS and I actually picked the PMOS device from Simulation_SPICE.
Not a problem, it is still a design schematic and has not been sent to be made.
Maybe this flags up that simulation components could be in a different colour (like FreeCad construction geometry vs normal geometry)?
Ha, you’re not the first one so don’t feel bad. Simplest solution is to just disable the SImulation_SPICE library under Preferences > Manage Symbol Libraries.
A good idea …
?stnenopmoc fo sepyt eseht rof esrever ni krow syawla I
Congratulations. No mistakes in your typing; I checked 
It is not so easy typing in reverse. I must admit, though, if the sentence is already displayed above the reverse version the job is not that difficult.
I was once asked to type a sentence in reverse with only the forward version in my mind. It took forever.
$ rev
?stnenopmoc fo sepyt eseht rof esrever ni krow syawla I
I always work in reverse for these types of components?
There’s a Linux app for everything, and for the rest you can program it.
