Patch 2 for RAD Studio 10.4 now available

[Marco Cantu 78]

Just FYI, the new (large) patch is out.
https://blog.marcocantu.com/blog/2020-july-radstudio104-patch2.html

 

over 20 Delphi compiler fixes, among a long list of items addressed. Fixed the debugger instability and a few more critical issues, also improved Delphi LSP quite a bit. Quality work will continue, this is a first step.

[Marco Cantu 78]

(and sorry if I don't follow up, I'm out a few days)

[Lars Fosdal 1791]

Have a good vacation and stay healthy. 

[Vincent Parrett 750]

10.4 Update 2 breaks code that compiled fine with update 1 and earlier

This is the library that shows the problem

https://github.com/VSoftTechnologies/VSoft.Messaging

 

  TChannelHelper = record
  public
    procedure PostMessage<T : record>(const message : T);
  end

This is my message record type

TGoodByeMessage = record
    MsgID  : TMessageID;
    Filler : TMessageFiller;
    Name   : string;
  public
    constructor Create(const theName : string);
  end;

It failed here with

 

[dcc32 Error] MsgDemoMainForm.pas(62): E2010 Incompatible types: 'T' and 'TGoodByeMessage'

 

procedure TForm2.goodbyeButtonClick(Sender: TObject);
var
  msg : TGoodByeMessage;
begin
  msg := TGoodByeMessage.Create('world');
  FChannel.Queue.PostMessage(msg); //<============
end;

 

This looks like records are broken with this updates, why? I'm guessing this has something to do with managed records?

 

[FredS 138]

3 minutes ago, Vincent Parrett said:

This looks like records are broken with this updates

How could they know you where going to do that 🙂

 

Sorry, couldn't help myself..

 

[pyscripter 689]

Also bugs reported as fixed are not fixed:

RSP-29271    [DelphiLSP] Code Insight adds unneeded () when changing procedures/functions

[Vincent Parrett 750]

More on my issue, removing the record constraint does allow it compile, but removes some compile time type safety  

[David Heffernan 2345]

40 minutes ago, Vincent Parrett said:

More on my issue, removing the record constraint does allow it compile, but removes some compile time type safety  

Will it compile if you restore the constraint but specify the generic type explicitly instead of relying of type inference? 

[Vincent Parrett 750]

13 minutes ago, David Heffernan said:

Will it compile if you restore the constraint but specify the generic type explicitly instead of relying of type inference? 

No, that was the first thing I tried, get this error instead

 

 [dcc32 Error] MsgDemoMainForm.pas(62): E2512 Type parameter 'T' must be a non-nullable value type

 

Which makes me suspect that record constraints are broken. My records have constructors, I'm guessing that now makes them managed records (I really haven't paid too much attention to how managed records work).

[David Heffernan 2345]

19 minutes ago, Vincent Parrett said:

My records have constructors, I'm guessing that now makes them managed records

No I don't think it does make them managed records. 

 

I don't really understand how such simple code could break the compiler. It's as if Embarcadero don't have comprehensive tests in place. 

[Remy Lebeau 1392]

19 minutes ago, Vincent Parrett said:

I really haven't paid too much attention to how managed records work

Custom Managed Records Coming in Delphi 10.3 (deferred to 10.4)

Custom Managed Records Coming to Delphi 10.4

Custom Managed Records New in Delphi 10.4 Sydney

Custom Managed Records

[Vincent Parrett 750]

4 minutes ago, David Heffernan said:

I don't really understand how such simple code could break the compiler. It's as if Embarcadero don't have comprehensive tests in place. 

My thoughts exactly. 

[Vincent Parrett 750]

Ok, so they are not managed records, the compiler is just broken.

[Vincent Parrett 750]

Logged as https://quality.embarcadero.com/browse/RSP-30073

Edited July 18, 2020 by Vincent Parrett

[Arnaud Bouchez 407]

This is the famous "patch release before holidays" syndrome...

🙂

 

 

[Guest]

@Arnaud Bouchez That video almost made me soil my pants, if that is the case, ( not sure can't stand yet ) then you owe me the cleaning service cost.

[Stéphane Wierzbicki 45]

1 hour ago, Arnaud Bouchez said:

This is the famous "patch release before holidays" syndrome...

🙂

 

 

Excellent! You make my day Arnaud

[dummzeuch 1505]

I'm sure that 'nobody got a raise for decade' is wrong. Nobody in IT got a raise, but surely management did. They handled the fallout caused by the mistakes of those stupid IT people after all!

Edited July 18, 2020 by dummzeuch

[Dalija Prasnikar 1396]

"If you can't laugh, what can you do?"

[Dalija Prasnikar 1396]

Honestly... I don't care if compiler broke your code... they released fix for compiler bug that broke my code... 

Edited July 18, 2020 by Dalija Prasnikar

[Vandrovnik 214]

Classic undocked still unusable...

How much time would it need to allow users to manually increase the height of the main window, so that icons are not hidden - 30 second? Or 60 with testing?

[dummzeuch 1505]

They managed to bungle the fix(es) for StringGrid custom drawing again. I just had to adp my Delphi 10.4 fix in GExperts for that. Did nobody test theses fixes?

[Stefan Glienke 2002]

@Vincent Parrett Ewww looks like they made record constraint only allow record with no managed fields although things were just working 

[Dalija Prasnikar 1396]

11 minutes ago, Stefan Glienke said:

@Vincent Parrett Ewww looks like they made record constraint only allow record with no managed fields although things were just working 

LOL, it looks like fix for my code literally broke Vincent's code https://quality.embarcadero.com/browse/RSP-28761

[Vincent Parrett 750]

2 hours ago, Dalija Prasnikar said:

LOL, it looks like fix for my code literally broke Vincent's code https://quality.embarcadero.com/browse/RSP-28761

ok I'm blaming @Dalija Prasnikar for logging bugs on such esoteric code that no sane person would ever use