BASIC

I think this does everything that was asked for:

10 PRINT "Hello World, from line 10!"

Java

public class Hello{
    public static void main(String[] args) {
        System.out.println("Hello World, from line "+new Exception().getStackTrace()[0].getLineNumber()+"!");
    }
}

Sinclair Basic

10 PRINT "Hello world, from line ";PEEK 23621;"!"

Hello world, from line 10!

This will work for any line by PEEKing at the address storing the current line number, so the following will work as well:

341 PRINT "Hello world, from line ";PEEK 23621;"!"

Hello world, from line 341!

Perl

close STDERR;
open FOOBAR,">",\$_;


print!warn,'Hello World, from line ',/(\d+)\.$/,'!';

Not quite as short as using __LINE__, but perhaps more interesting.

warn is a debugging tool, which issues a statement to STDERR, indicating in which file, and on which line the warning was issued... unless STDERR has been previously closed or is otherwise inaccessible, in which case the warning is issued to the most recently opened file handle - this is undocumented behavior. I'm not sure if it would be better classified as a feature or a bug.

Here, STDERR is closed, and a new file handle identified as FOOBAR is opened, and routed to the variable $_. This is then parsed to retrieve the line number of the warning, which is embedded in the print statement.

C

#include <stdio.h>
main(){
printf("Hello World, from line %d!", __LINE__);
}

Lua

print("Hello world, from line "..debug.getinfo(1).currentline.."!")

Python

Example (10 lines, 213 characters):

import sys
import traceback
lineno = None
while True:
    try:
        print 'Hello World, from line %d!' % lineno
        break
    except:
        lineno = traceback.tb_lineno(sys.exc_info()[2])
        continue

Try online here. Non-flow altering code and whitespace can be added and the program will display the updated line count, and likewise, this code snippet can also be used anywhere in an existing program. Expected output:

Hello World, from line 6!

Another example (try online here) to show that it works when code/whitespace is added. Expected output:

Down we go...
Gotta catch 'em all.
Down we go...
Hello World, from line 11!
Awesome!

Python 3

import hashlib
with open(__file__) as f:
    line_num = 0
    for line in f.readlines():
        line_num += 1
        if hashlib.sha256(bytes(line, "UTF-8")).hexdigest()[0:6] == 'cc46f7':
            print('Hello world, from line {}!'.format(line_num)) # cc46f7

Hello world, from line 7!

This self-reading code contains a self-referencing hash. The SHA256 sum of the last line begins with cc46f7.... When it hashes the print line, it finds that the hash matches the magic value it's searching for.

Javascript

function getLine(n) {
   try {
      to
   } catch (dat) {
      var stack = dat.stack.split('\n');
       for (var i = 0; i < stack.length; i++) {
           if (~stack[i].indexOf ('getLine')) break;          
       }
      return dat.stack.split ('\n')[i + ~~n].match (/:(\d+)/)[1] - ~~window.hasOwnProperty ('__commandLineAPI')
   }
}
//Line 12
console.log ('Hello World, from line ' + getLine(1) + ' !')

Note: Expressions, evaluated from within the chrome dev console will be wrapped in a with statement. Thus we need to decrement the line by one if that's the case

C++

#include <iostream>
#include <utility>
#include <type_traits>

#define __(A,B,C,...) B##C##A
#define _(A,...) __(__VA_ARGS__, A)
template<unsigned...Is> struct v;
template<unsigned I0, unsigned... Is> struct v<I0, Is...>:v<I0-1,I0-1, Is...> {};
template<unsigned...Is> struct v<0,Is...>:std::integral_constant<unsigned, sizeof...(Is)> {};

int main() {
  std::cout << "Hello world from line " << v<_(EXTRACT,_(Q,_,E,_,$$main$$,),_(@22,,_,_),_(Z,N,L,I,_,,L),__STACK_TRACE__)>::value << "\n";
}

live example

Python

import traceback, inspect
frame = inspect.currentframe()
print("Hello World, from line "+traceback.format_stack(frame)[0].split()[3][:-1]+"!")  

Bash

#
# some comments to fill some lines...
#
echo "Hello World, from line $LINENO!"

Output

Hello World, from line 4!

D

void main ()
{
    import std.stdio;
    writefln("Hello World, from line %d", __LINE__);
}

Perl

print 'Hello World, from line '.__LINE__.'!';

Java

public class HelloFrom {
    public static void main(String[] args) {
        System.out.println("Hello World, from line " + Thread.currentThread().getStackTrace()[1].getLineNumber() + "!");
    }
}

C#

C# 5.0's [CallerLineNumber] does the trick:

using System;
using System.Linq;
using System.Runtime.CompilerServices;
namespace LineNumberConsole
{
    class Program
    {
        public static void Main()
        {
            Console.WriteLine("Hello World, from line {0}!", ToRoman(GetLineNumber()));
            Console.ReadLine();
        }

        private static int GetLineNumber([CallerLineNumber] int sourceLineNumber = 0)
        {
            return sourceLineNumber;
        }

        private static string ToRoman(int number)
        {
            // TODO: Copy some source code from http://pastebin.com/w0hm9n5W
            // Skipped here for brevity
            return number.ToString();
        }
    }
}

Output

Hello World, from line X!

C

By using a variadic macro, we can make a print function which automatically adds the line number to the end of an arbitrary printf statement, and always prints to stdout.

test.c:

#include <stdio.h>

#define printfl(format, ...) fprintf(stdout, format " From line %d\n", ##__VA_ARGS__, __LINE__)

int main() {
    printfl("Hello World! I have %d argument(s).", 1);
    return 0;
}

outputs:

% ./test
Hello World! I have 1 argument(s). From line 6

Note: I deviated from the pattern to demonstrate that printfl is still a valid variadic function; if you really care about the format of the output, you can always change the literals I use.

Java

Using the behavior of Exception's stack trace to get current line. as long as Printstatement isn't mangled into multiple lines or classfile gets mangled it should work

public class PrittLnbr
{
    public static void main(String[] args)
    {
        System.out.println("Hello World, from line "+new Error().getStackTrace()[0].toString().split(":")[1]+"!");
    }
}

Kind of boring in Ruby:

puts "Hello World, from line #{__LINE__}!"

This isn't cheating, right?

Python

import inspect
print ("Hello world from line %d!" % (inspect.getlineno(inspect.currentframe())))

Output

Hello World from line 2!

C or C++, and AWK

lineno.c:

// code or comments
// ....
#error Hello World, from line
// other code or comments

Usage:

gcc lineno.c 2>&1 | awk '{ split($0,a,":"); ; printf("%s %s!\n", gensub(".*#error ","",1), a[2]); exit; }'

Output:

Hello, World, from line 3

Notes:

• No user written code searches the file.
• g++ will work on a c++ file.

Inspired by @squeamish ossifrage's answer with BASIC.

What's in a line number?

In, for instance, IBM COBOL there are (can be) arguably three source line numbers.

The first (if used, it is optional), in columns 1-6 of the punched-card, are sequence numbers for the card deck. If the deck is dropped, the otherwise-hapless programmer has some chance to get things back in order. OK, starts to go slightly floppy as soon as the program has lines added, but that's what different coloured cards are (can be) for.

The second is on the compiler listing, generated by the compiler. The compiler diagnostic messages use this line number, else no-one would have much of a clue (unless the diagnostics are embedded, which they can be) which line any particular diagnostic referred to*. These numbers for various reason (copybook code, let alone anything else) will be different from the first.

The third line number is of course the line number of the actual source file (optional, columns 73-80 if present). Other than actually reading the source file, or hard-coding a reference to a particular source line-number, it would not be possible to get this number in a program***.

So:

999999     DISPLAY "Hello, world, from line 999999!"

This can appear anywhere in the PROCEDURE DIVISION of the program, since those line numbers have no affect on the program, and are only even flagged as out of sequence (not ascending, can be gaps) if requested by compiler option.

** With compiler options SEQUENCE and NUMBER, the column 1-6 numbers will be used (modified by the compiler if necessary) for the diagnostic messages, and other data requiring source line numbers.

* OK, possible in limited circumstances. No copybooks. Then you get the editor to put the sequence numbers of the source program (yes, we have actual sequence numbers, not just counting the lines) in columns one to six, and use compiler options SEQUENCE and NUMBER and then have a DEBUG DECLARATIVE, have the DISPLAY on the same physical line as the paragraph/SECTION which is going to cause the DEBUG DECLARATIVE to execute.

Cobra

class Program
    def main
        print 'Hello World, from line [System.Diagnostics.StackFrame(true).getFileLineNumber]!'

Ruby

File.write "hello.rb", "x=2\n"+"x+=1\n"*rand(rand(100))+'puts "Hello World, from line #{x}!"'
system "ruby hello.rb"
File.delete "hello.rb"

Befunge

Done just for fun.

>00g1+:00p"v"\10v  
    v-*45g00p\g <  
#v+1_$10g1vv,,,,<  
^<p000p01+<#>,,,^  
>" enil morf ,oll"v
@.,,,,,,,,,<^,"He"<

Conditional: top left of the code has to be 0 < x < 20 and 0 <= y < 62; and the two first cells have to be empty.

example:

                         v                  

                         0                  
                         0                  
                         0                  
                         p                  
                         0                  
                         1                  
                         0                  
                         p                  
                         >00g1+:00p"v"\10v  
                             v-*45g00p\g <  
                         #v+1_$10g1vv,,,,<  
                         ^<p000p01+<#>,,,^  
                         >" enil morf ,oll"v
                         @.,,,,,,,,,<^,"He"<

Would output:

Hello, from line 10

GNU COBOL

Well, they said it couldn't be done. Actually, it was I who said it couldn't be done. Now it's done, and an obsolete language feature re-implemented using the method applied.

The question states:

If any additional whitespace or statements (which do not interrupt the flow of the code) is added to the source code, it should be reflected at run-time (after compiling if applicable).

Any amount of stuff can be inserted prior to the three DISPLAYs which cause the start of the output, and anything after the DISPLAYs would "interrupt the flow of the code", so that's OK.

COBOL used to have a TRACE verb (statement) which simply listed source line-numbers as they were executed (no access to the line number in the program). Although of limited use, I've included an implementation of TRACE.

   ID Division.
   Program-ID. HIWHERE.
   ENVIRONMENT DIVISION.
   configuration section.
          source-computer. TinkerToy with debugging mode.
   Procedure Division.
   Declaratives.
   Debug-Declaratives Section.
       Use For Debugging on a b
       .
   Debug-Declaratives-Paragraph.
       Display Debug-Line "!"
       .
   End Declaratives
       .
   Main-Program Section.
       DISPLAY "Perform"
       Display "Hello World, from line " no advancing Perform b
       display "GO TO"
       Display "Hello World, from line " no advancing GO TO a
       .
   a.
       dISPLay "Fall through"
       Display "Hello World, from line " no advancing. b.
   The-Last-bit-OF-the-PROGRAM.
       GOBACK
       .

The output is

Perform
Hello World, from line     18!
GO TO
Hello World, from line     20!
Fall through
Hello World, from line     23!

As an exhibition of the power and flexibility of writing the language, this example uses mixed-case, entirely lowercase, and entirely uppercase, all at the same time. It does not matter, as when processed, everything is "folded" to UPPERCASE.

The only standard COBOL way to get at a source line-number in the running program, from the running program, is with a DEBUGGING DECLARATIVE. Within a SECTION, strictly within a paragraph within a SECTION, of a such a declarative you have access to the special-register DEBUG-LINE. This contains the source line-number of the verb (statement) which caused transfer of control to a particular procedure-name (paragraph or SECTION).

So, with PERFORM, or GO TO, or "fall through" the paragraph in the debugging declaratives SECTION is executed.

OK, but DISPLAY does not cause transfer of control.

No problem. Put it on the same line as the transfer of control.

Problem, since if "any additional whitespace or statements (which do not interrupt the flow of the code) is added to the source code, it should be reflected at run-time (after compiling if applicable)".

So, put it on the same line but in front of a transfer of control, split the content of the DISPLAY into two pieces (remember, "In this context, we want the first line number of the statement which outputs the string to be displayed") and output the first part prior to the transfer of control, and the second part, from the DEBUG-LINE, once inside the debugging procedure.

The final tricky bit is for the "fall through" ("procedures" can be PERFORMed, can be the target of a GO TO, or can be entered simply by being the next line along). In this instance, put the DISPLAY on the line which defines the procedure, but in front of the definition.

The names of the "procedures" (a and b) have been severely shortened to allow them to fit on the same source-line as the DISPLAY. Strictly a COBOL procedure-name should start somewhere from column eight to column 11. However, the syntax is, these days, much more relaxed about that. To the extent that I can define a procedure name on the same line as some code. Even embedded in code. Care, and an occasional full-stop, is required.

In the PROCEDURE DIVISION each full-stop shown is required, and no more are.

To compile:

cobc -x -g hiwhere.cbl

To execute (linux):

COB_SET_DEBUG=Y ./hiwhere

Finally, the return of TRACE (without READY/RESET).

   ID Division.
   Program-ID. tRacE.
   ENVIRONMENT DIVISION.
   configuration section.
          source-computer. TinkerToy with debugging mode.
   Procedure Division.
   Declaratives.
   Debug-Declaratives Section.
       Use For Debugging on a
       .
   Debug-Declaratives-Paragraph.
       Display Debug-Line
       .
   End Declaratives
       .
   Main-Program Section.
  *    Just append "perform a" to a single-line statement.
       DISPLAY "1" . perform a
       Display "2" . perform a
       display "3" . perform a
  *    Or prepend "perform a." for a multi-line statement, or a
  *    statement which won't "come back". 
       perform a. GOBACK
       .
   a.
       CONTINUE
       .

Output is:

1
    17
2
    18
3
    19
    20

Where 1, 2 and 3 are output from the three DISPLAY statements, and 17, 18, 19 and 20 are the line numbers of the "executable" (non-debugging) lines.

Powershell

$l=(Get-PSCallStack | ForEach{$_.Location})[0].split(' ')[-1]; "Hello World, from line $l!"

And:

try{ I AM ERROR. } catch { $l=$error[0].InvocationInfo.ScriptLineNumber; "Hello World, from line $l!" }

Both work like this:

PS C:\MyFolder> .\helloworld.ps1
Hello World, from line 1!

Python

import traceback

print 'Hello World, from line %i!' % traceback.extract_stack()[0][1]

Short and sweet.

Javascript

One line using stack trace.

(function (o) { console.log("Hello World, from line " + (Error.captureStackTrace(o) || o.stack.match(/\d+/)[0] - !!__commandLineAPI) + "!"); })({});

PowerShell

Cheap Move

Function LemmeGetDatError() {
    "Too busy chuggin along"
    "Then all of a sudden, I meet a new programmer"
    "And he's all like"
    Write-Output "$(Try {"Hello World from"; Throw "error" } Catch {$_.ScriptStackTrace.Split(":")[1]})"
}

LemmeGetDatError