Understanding the {$WARN XXXX OFF} compiler directive - freepascal

Question

If I'm trying to indiscriminately silence compiler messages the compiler command line parameter -vmXXXX works perfectly for the whole project. However, if I'm trying to silence messages selectively in the source I have to use the {$WARN XXXX OFF} compiler directive.
The problem is... I am having a really hard time understanding the rules of how that compiler directive is applied. In the example below of 1 program and 2 units, if you uncomment the {$WARN XXXX OFF} command one at a time, what will happen is not what I would expect (the result is written in the comment for each compiler directive).
Is there anyone who can explain exactly how that compiler directive works?
Program:
program Project1;
//{$WARN 4055 OFF} // will silence the unit2 hint but not the unit1 hint
uses
Unit1;
//{$WARN 4055 OFF} // will silence neither the unit2 hint nor the unit1 hint
begin
if convert_it(0)=nil then
halt(1);
end.
First Unit:
unit Unit1;
//{$WARN 4055 OFF} // will silence the unit2 hint but not the unit1 hint
interface
//{$WARN 4055 OFF} // will silence the unit2 hint but not the unit1 hint
function convert_it(avalue:longint):pointer;
//{$WARN 4055 OFF} // will silence the unit2 hint but not the unit1 hint
implementation
//{$WARN 4055 OFF} // will silence the unit2 hint but not the unit1 hint
uses
unit2;
//{$WARN 4055 OFF} // will silence the unit1 hint but not the unit2 hint
function convert_it(avalue:longint):pointer;
begin
result:=pointer(do_it(avalue));
end;
end.
Second Unit:
unit Unit2;
//{$WARN 4055 OFF} // will silence the unit2 hint but not the unit1 hint
interface
//{$WARN 4055 OFF} // will silence the unit2 hint but not the unit1 hint
function do_it(avalue:longint):longint;
//{$WARN 4055 OFF} // will silence the unit2 hint but not the unit1 hint
implementation
//{$WARN 4055 OFF} // will silence the unit2 hint but not the unit1 hint
function do_it(avalue:longint):longint;
var
p:pointer;
begin
p:=pointer(avalue);
if p=nil then
result:=avalue+1
else
result:=avalue-1;
end;
end.

Answer 1

From the manual:
Local directives can occur more than once in a unit or program, If
they have a command line counterpart, the command line argument is
restored as the default for each compiled file. The local directives
influence the compiler’s behaviour from the moment they’re encountered
until the moment another switch annihilates their behaviour, or the
end of the current unit or program is reached.
So the directives you use are valid from their declaration to the end of a unit, following the use clauses, after what the default value is restored.
For each unit, for each nth directive
Program
1: the directive is activated, still valid in unit1 because of the use, still valid in unit2 because of the use. But after unit2 (end of a unit) the default value is restored, so you got the message in unit1.
2: the directive is activated after the use so it has no effect whatever is the unit.
Unit1
1: the directive is activated, still valid in unit2 because of the use. But after unit2 (end of a unit) the default value is restored, so you got the message in unit1.
2: ditto
3: ditto
4: ditto
5: the directive is only activated for unit1 implementation section.
Unit2, all cases: the directive is only activated for unit2.
The tricky thing to get here is that it follows the use.