likely bug in comment/directive parsing

[440bx]

Hello,

Consider the following code:

Code: Pascal  [Select][+][-]

1. {$MODESWITCH NESTEDCOMMENTS OFF }
2.  
3. program _Sizeof;
4.  
5.  
6. type
7.   ANYSIZE_ARRAY = 0..0;
8.  
9. type
10.   { _BOOT_ENTRY                                                               }
11.  
12.   PBOOT_ENTRY = ^TBOOT_ENTRY;
13.   TBOOT_ENTRY = packed record
14.     Version              : DWORD;
15.     Length               : DWORD;
16.     Id                   : DWORD;
17.     Attributes           : DWORD;
18.     FriendlyNameOffset   : DWORD;
19.     BootFilePathOffset   : DWORD;
20.     OsOptionsLength      : DWORD;
21.  
22.     OsOptions            : packed array[ANYSIZE_ARRAY]  { OsOptionsLength     }
23.                                      of byte;
24.   end;
25.  
26.   { unterminated comment - CLOSING BRACE MISSING
27.  
28.   {$if sizeof(TBOOT_ENTRY) <> $1d}
29.     {$FATAL sizeof TBOOT_ENTRY does not equal $1d}
30.   {$endif}
31.  
32.  
33. begin
34.   writeln('sizeof TBOOT_ENTRY: ', sizeof(TBOOT_ENTRY));
35.  
36.   readln;
37. end.    
38.  

There is a closing brace missing in line 26.

When attempting to compile the program, the compiler emits the message "sizeof TBOOT_ENTRY does not equal $1d" indicating that the preprocessor found the $FATAL message in spite of it being in a regular comment (the unclosed comment.)

Additionally, in spite of the NESTEDCOMMENTS OFF directive, there is no message/warning/error about the nested comment.

[Martin_fr]

There is a closing brace missing in line 26.

When attempting to compile the program, the compiler emits the message "sizeof TBOOT_ENTRY does not equal $1d" indicating that the preprocessor found the $FATAL message in spite of it being in a regular comment (the unclosed comment.)

Additionally, in spite of the NESTEDCOMMENTS OFF directive, there is no message/warning/error about the nested comment.

It's late, but...

If nested comments are OFF, then the comment starting in line 26 ends in line 28 (at the end of "{$IF....").
The extra opening "{" does not open a nested comment, so a simple single level comment keeps going....

Then of course when the comment ends at eol line 28, the {$FATAL on line 29 is seen?

[Khrys]

Additionally, in spite of the NESTEDCOMMENTS OFF directive, there is no message/warning/error about the nested comment.

Like already @Martin_fr explained, this is to be expected -  {$modeswitch nestedcomments off}  turns off support for nested comments at the parser level (as opposed to still parsing them & generating fatal errors).
Turning on nested comments also behaves exactly as expected - the  {$if}  directive is assumed to be a nested comment (and is reported as a warning) while the comment continues until EOF.

[Thaddy]

It is more like the editor at hand (Lazarus?) confuses you by syntax highlighting code that is in effect a comment so should be greyed/ comment styled when nested comments are off.
This also happens - or does not happen! - with the officially documented comment style {$if 0}/{$ifend}

Basically it is a shortcoming of the IDE. The compiler handles it correctly.

[440bx]

I understand the thinking but, I find it rather dubious.   The opening brace is not a closing brace (and should not be thought of as one) therefore it cannot be considered as closing the unclosed comment.

The absence of a closing brace creates a syntactically incorrect construction the compiler should simply reject.

The compiler has been specifically told that there should not be nested comments yet, it ignores the nested comment, i.e, the sequence of two opening braces, which should result in a compiler error.

NESTEDCOMMENTS OFF, that's just a xmas ornament!.  It tells the compiler that a second opening brace acts as a implied closing brace.

[jamie]

@440ex
Off topic:
 Are you sure the last field is not supposed to be 0 size content?

Jamie

[paweld]

@440bx: https://forum.lazarus.freepascal.org/index.php/topic,54458.msg404554.html#msg404554

[Thaddy]

@440bx
The compiler won't even see it....Why should it parse commented code?

[440bx]

I consider FPC's entire handling of block comments to be erroneous but, I realize it's not going to change regardless of the resulting atrocities.

Obviously, the preprocessor which is part of the scanner parses some comments.  Since the scanner is part of the compiler, it is a fact that the compiler (a specific part of it) sees it.

I'm done with this.

[Martin_fr]

The compiler has been specifically told that there should not be nested comments yet, it ignores the nested comment, i.e, the sequence of two opening braces, which should result in a compiler error.

NESTEDCOMMENTS OFF, that's just a xmas ornament!.  It tells the compiler that a second opening brace acts as a implied closing brace.

NESTEDCOMMENTS OFF does not mean to give an error on nested comments.

It means not to "know about the possibility of nesting" (and then consequently, they don't exist, so they can not give an error).

Code: Pascal  [Select][+][-]

1. {$MODESWITCH NESTEDCOMMENTS OFF }
2.  
3.  { start a comment
4.   ....
5.  { this is just another commented char. It has no meaning at all.
6.   ....
7.   }  // this closed the comment
8.  
9.  
10.   // < this starts a comment // < this does not start a nested comment
11.  

Inside a comment, any further "{" is not seen as start of a (nested) comment. Its no different than any other char (except the closing "}").

[440bx]

The open brace in Pascal is a special character that indicates the start of a block comment. 

Whenever an open brace is found, the compiler should interpret that as an attempt of opening a comment (that's what the open brace is for), therefore an open brace found inside a comment should be seen as an attempt of creating a nested comment, which is something not supported (at least not in standard Pascal.)

The attempt of supporting nested comments is a mistake with highly undesirable consequences.

[Thaddy]

@440bx

THAT is something I wholly agree with.
Rip it out, nobody uses it or relies on it anyway.

I also never found any use for it, looks more like a hobby project.

Code: Pascal  [Select][+][-]

1. {$but}

maybe?

[Martin_fr]

The open brace in Pascal is a special character that indicates the start of a block comment. 

Whenever an open brace is found, the compiler should interpret that as an attempt of opening a comment (that's what the open brace is for), therefore an open brace found inside a comment should be seen as an attempt of creating a nested comment, which is something not supported (at least not in standard Pascal.)

The attempt of supporting nested comments is a mistake with highly undesirable consequences.

In a string too? SCNR

But joke aside, it's a matter of definition.

$NESTEDCOMMENT is defined as making it (or making it NOT) such a special char.

What should
$NESTEDCOMMENT off
do, if not "making it a normal char inside comments"?

If it keeps it a special char, then when found inside a comment, it would be a nested comment. But that conflicts with having that turned off... So what then? An error? Well the mode for "an error" just does not exist.

[440bx]

The presence of a { inside a comment that was started with a { should _always_ be considered an error.

[Thaddy]

Only with nestedcomments on.
I don't get it: the parser knows it need to find the closing. IT DOES!
It should not even see it if nestedcomments are off, so where is your problem?

But as I wrote: who the hell allowed nested comments anyway.

Petty project?