
Each directory a, b, c, etc contains a test case with these files:

environment
	common environment file

environment.$context
	context-specific environment file (happens after common file)

output.irafdev		<- output for context=irafdev on any machine
output.irafx.gaudete	<- output for context=irafx on gaudete
output.irafx.thor	<- output for context=irafx on thor
	output of dictionary created for that directory, similar to
	the output of /usr/bin/env


If you don't see the .irafx. files, there is no file for that context in that tree.




The grammar is something like this.  Use and/or modify as convenient.

* block_name		<- start of block, name has no spaces
var=value		<- variable assignment; to end of line for now
var=value
;			<- end of block.  use it or ignore it


* next_block_name	<- start of next block
var=value
;


TOPLEVEL		<- if the environment file contains this line, it
			   is at the top level of the directory hierarchy


blank lines ignored
# is a comment

blocks are processed in the order you find them
vars are processed in the order you find them

the name of the block can be 
	"all" to apply everywhere
	"host=xyz" to apply on host named xyz
		host name is platform.node().split(".")[0]

( let's say we'll think about generalizing this to os, cpu, or whatever
if necessary)

( let's say we'll think about "host=xyz|abc" at some point.)


b.t.w.  We will want a way to say "get all the environment information from
above here" so you can pick a subdirectory out of the tree to run it yourself.
It might insert PDK_TESTPREFIX into the environment it reports.

