bash is not in the base system, so it will not be found in PATH unless shells/bash is installed.

These should go before the options block. See Chapter 15. Order of Variables in Port Makefiles.

I thought this line caused it to be installed? Could you check the changes I've made to see if they're better/worse please?

The tag says ...stable201903 which looks a lot like "march 2019", is there a reason the version up there says april 2019 and not march?

All those are out of order. See Chapter 15. Order of Variables in Port Makefiles.

Order is :

1. USES, USE_* and related variables
2. Standard framework variables, like MAKE_ARGS
3. Options, and then options helpers
4. Variables local to the Makefile

I know, it _should_ be 20190307, the date the tag was added - but I guess it's dated for when the port was updated. I'd prefer not to change it now, but instead wait until I update it to the next edk2-stable201911 tag.

Are there slave ports using this one as a master port ?

If not, don't use +=.

This is in the USES/USE_* block, should be just after USES.

This is a standard framework variable, goes before options, with MAKE_ARGS.

-j4 is so wrong, a port should not force or decide which number of build jobs it is going to use. This is set by the user, globally, and is available through MAKE_JOBS_NUMBER.

You should use -j${MAKE_JOBS_NUMBER}.

Also, technically, you should not be using MAKE_CMD directly, but DO_MAKE_BUILD, probably something along the lines of:

${DO_MAKE_BUILD} -C ${WRKSRC}/BaseTools

I have not checked, but it feels like _GCC5 is strange, we don't have gcc5 in the tree, and the default is 8 or 9 depending on the arch.

It looks like I can't use ${DO_MAKE_BUILD} because it uses its own environment and won't work with the Makefile that EDK2 uses: for example EDK2 invokes "make" within the Makefiles.

That's the EDK2 toolchain, which supports everything from GCC 5 and newer with the same "GCC5" toolchain definition.

Why did you bump PORTEPOCH?

Because otherwise portlint said the version was going backwards. That is, that "0.2" is a higher version number than "g20190424".
It said I should either bump PORTEPOCH or there was another variable I could use (PORTREVISION?).

OK, PORTEPOCH is ok then.