This is a web server problem or configuration issue.  Not a PHP issue.

As evidently shown, PHP accepted the original request uri of 
/test.txt/fake,php, and evidently shown in the php_info , it took that 
and changed the script_file name to test.txt, 

It should have tried to execute fake.php and returned no file could be 
found. If the webserver had instead sent test.txt as the 
script_filename, and /fake.php as the path_info, then I could understand 
it happening, but it did not.

Well, first, there is no fake.php in your output there.  So I don't 
know about your "evidently" since your evidence doesn't match at all.  
And second, PHP doesn't set SCRIPT_FILENAME, your web server does.  So 
your web server determined that the script to execute was 
"/opt/html/domain/test.txt" evidently as per your own evidence.  So PHP 
executed that script.  I am not sure why that is surprising to you.  
Why is your web server telling PHP to run test.txt?  PHP doesn't check 
the filetype.  It runs what it is told to run.

1.php was from a previous test when I typed /1.php after test.txt, if 
you type /fake.php after test.txt it'll show fake.txt, just like how I 
don't own domain.com, you apparently can't go back and edit the 
original.

Regarding the webserver, Nginx matches up php via regex, as per the 
wiki provided 

http://wiki.nginx.org/PHPFcgiExample#Connecting_Nginx_to_the_running_F
astCGI_Process

location ~ \.php$ { ... } would match up to a php file request. 

It is my opinion that the PHP parser should take script_filename 
litterally. 

That is to say
script_filename = /test.txt/1.php
path_info = no value

shouldn't execute

but
script_filename = /test.txt
path_info = /1.php

should execute, if the webserver indeed sent it those values. In my 
case it only sent /test.txt/1.php , which should not have worked 
because that filename does not exist on the system.

It still doesn't matter.  By the time the request makes it to PHP, PHP 
is going to execute.  I bet you can't reproduce this on a properly 
configured Apache server that correctly finds the script and the 
pathinfo.  Note that in your output your ORIG_PATH_INFO is empty, which 
is incorrect.  If your URI is /test.txt/1.php and /test.txt exists, 
then "/1.php" is the PATH_INFO.  Obviously your server doesn't seem to 
understand that and determines the URI should be executed by PHP.  The 
executable part of your URI is /test.txt and it isn't PHP's job to 
second-guess that.

But PHP is second guessing, it was sent "/test.txt/1.php" as the 
original script_filename, that file doesn't exist on the system, so how 
does it execute the code if 1.php does not exist? As a result it was the 
parser that move down to the test.txt. Again, note the original 
script_filename sent to the parser.

PHP walks back through the script_filename it is passed and tries to 
find the script to execute.  Once again, it is the web server's job to 
figure out that a URI doesn't exist.  PHP doesn't serve up 404s under 
any circumstances.  If all your web server is doing is a simple regex 
to determine whether something should be executed or not, then your web 
server is broken.

Digging into it quote a bit for future readers of the bug. 

The ~ \.php { } matchup as taught by nginx wiki and nginx creator is 
indeed insecure if php is to execute the way described above. 

Turns out to fix, needs to have something like this:

location ~ [^/.][a-zA-Z0-9_-]+\.php[s]? { } 

doing such will 404 on the above /test.txt/fake.php example, but will 
still correctly parse path_info. 

#50837 CLOSED/BOGUS , no further argument from me. I'll go update the 
wiki