activex.pp definition problem

[440bx]

Hello,

In activex.pp there are a large number of GUID definitions that are supposed to be constant but, they are not.

for instance, it is possible to do this:

Code: Pascal  [Select][+][-]

1.   GUID_NULL.D1            := 1;
2.   IID_IEnumConnections.D1 := 1;
3.  

Known GUIDs should not be changeable.

[Thaddy]

Also the case in C++ btw... not a Pascal issue.
I guess you can declare these GUIDS as typed const, but in {$J-} state which makes them immutable.
I thought you knew that:

Code: Pascal  [Select][+][-]

1. {$push}{$J-}
2. const
3.   myguid:TGUid = '{77DE903C-1453-4622-93A0-1C46B74627B9}';
4. {$pop}
5. begin
6.   myguid.d1 := 5;  // Error: Can't assign values to const variable.
7. end.

It should be fine to compile the whole interface section of the activex unit in forced {$J-} state.
(If that is not already the case)
The guid ends up in .rodata and so is immutable:

Code: ASM  [Select][+][-]

1. .section .rodata.n_TC_$P$PROGRAM_$$_MYGUID,"aw"
2.         .balign 4
3. .globl  TC_$P$PROGRAM_$$_MYGUID
4. TC_$P$PROGRAM_$$_MYGUID:
5.         .long   2011074620
6.         .short  5203,17954
7.         .byte   147,160,28,70,183,70,39,185  # same guid, but in decimal.
8. # End asmlist al_rotypedconsts

Maybe @Marcov can check if this is the case and otherwise protect the interface section with {$J-} ?
This should not otherwise affect the code. Must be done with {$push}/{$pop}, though.

[440bx]

I thought you knew that:

You actually thought ?... I hope you didn't experience too much pain.

[Thaddy]

Nah... Hope its solves the problem, though.

I did:
- around line 101 {$PUSH}{$WRITEABLECONST OFF} // start of GUID consts
- around line 189 {$POP} // end of GUID consts, alternatively around line 4954 for the pop. Just above implementation.
That seems enough. Not really good with patching with GIT yet.
push/pop is necessary to avoid settings spill. 101/189 seems the most "correct": least code affected.

I will leave that up to Marcov.
Just in case:

Code: Pascal  [Select][+][-]

1. packages/winunits-base/src/activex.pp | 4 ++--
2.  1 file changed, 2 insertions(+), 2 deletions(-)
3.  
4. diff --git a/packages/winunits-base/src/activex.pp b/packages/winunits-base/src/activex.pp
5. index 56dec102fe..2420a7e5bc 100644
6. --- a/packages/winunits-base/src/activex.pp
7. +++ b/packages/winunits-base/src/activex.pp
8. @@ -98,7 +98,7 @@
9.     TOleColor           = OLE_COLOR;
10.     POleColor           = LPOle_Color;
11.     HHandle             = UINT_PTR;
12. -
13. +{$PUSH}{$WRITABLECONST OFF}
14.  CONST
15.     GUID_NULL  : TGUID =  '{00000000-0000-0000-0000-000000000000}';
16.     IID_IPrint : TGUID = '{B722BCC9-4E68-101B-A2BC-00AA00404770}';
17. @@ -186,7 +186,7 @@
18.     IID_IOleCache2 : TGUID = '{00000128-0000-0000-C000-000000000046}';
19.     IID_IOleCacheControl : TGUID = '{00000129-0000-0000-C000-000000000046}';
20.     IID_IOleItemContainer : TGUID = '{0000011C-0000-0000-C000-000000000046}';
21. -
22. +{$POP}
23.  
24.       // bit flags for IExternalConnection
25.  CONST
26.  

Should be close to a correct patch. (against trunk from today 10:00)

[nanobit]

Unfortunately {$J+} is default in FPC,
whereas Delphi has the preferable default {$J-} since Delphi 6.

[Thaddy]

See the bug report..... <sigh> And your statement about Delphi is not true D7 behaves the same, D12.1 behaves the same.

[nanobit]

Excerpt from Delphi 6 Developer’s Guide (Steve Teixeira):

Migrating from Delphi 5: Although compatibility between Delphi 5 and 6 is quite good, there are a few minor issues you should be aware of as you make the move.
Writable Typed Constants
The default state of the $J compiler switch (also known as $WRITEABLECONST) is now OFF,
where it was ON in previous versions. This means that attempts to assign to typed constants will raise a compiler error unless you explicitly enable this behavior using $J+.

[gues1]

In "Delphi 7 Object Pascal Language Guide" at page 5-43:

In the default {$J-} compiler state, typed constants can not have new values assigned
to them; they are, in effect, read-only variables.

[Thaddy]

Yes. Hence my patch.