You RAD Studio 10.4 Sydney appreciated features and bug fixes

[Wagner Landgraf 43]

Share here what you liked the most about 10.4 Sydney, be it new features or bug fixes you were waiting for.

For issues and complains, use this topic: https://en.delphipraxis.net/topic/2885-your-rad-studio-104-sydney-issues/

[Lars Fosdal 1792]

The new Error Insight is GREAT! It works really well!

After years of red squigglies for things that are not wrong, it is really nice to only see squigglies when something actually is wrong.

[Dalija Prasnikar 1396]

This one is rather important fix - it is fixed for both LSP and Classic Code Insight

 

Code completion does not work inside anonymous methods

https://quality.embarcadero.com/browse/RSP-23252

[pyscripter 689]

Right now the quality portal is updated at a frenetic rate.  There was a very large number of fixes, I think more than in any other previous version, and I was quite pleased that quite a few of my reports have been resolved, some of the unexpectedly:

• System.AnsiStrings AdjustLineBreaks not exported
• Memory leak in TInvokeableVariantType.DispInvoke
• Optimize TInvokeableVariantType.DispInvoke
• Wrongly premultiplied TWICImage when assigning a Bitmap with aDefined
• TButtonedEdit is not styled properly

 

System.Threading got some attention with the famous Cancel and Wait issue resolved.

 

Issues I would like to see fixed now and consider critical are:

 

• TThreadedQueue and TMonitor issue, possible solution
• InterlockedCompareExchange128 doesn't restore RBX
• Threading - Incorrect calculation of IdleWorkerThreadCount 

 

 

Edited May 28, 2020 by pyscripter

[pyscripter 689]

Also the improvements in Vcl.Styles and DPI awareness are welcome.   There are still a few outstanding issues:

 

• InputQuery scaling and styling issues
• Styled menus DPI scaling
• TCustomizeDlg not fully VCL styled

 

 

Edited May 28, 2020 by pyscripter

[Tommi Prami 130]

I am pretty interested in hearing what are people thinking of implementing with Managed records.

 

-Tee-

[David Heffernan 2345]

6 hours ago, pyscripter said:

Right now the quality portal is updated at a frenetic rate.  There was a very large number of fixes, I think more than in any other previous version, and I was quite pleased that quite a few of my reports have been resolved, some of the unexpectedly:

• System.AnsiStrings AdjustLineBreaks not exported
• Memory leak in TInvokeableVariantType.DispInvoke
• Optimize TInvokeableVariantType.DispInvoke
• Wrongly premultiplied TWICImage when assigning a Bitmap with aDefined
• TButtonedEdit is not styled properly

 

System.Threading got some attention with the famous Cancel and Wait issue resolved.

 

Issues I would like to see fixed now and consider critical are:

 

• TThreadedQueue and TMonitor issue, possible solution
• InterlockedCompareExchange128 doesn't restore RBX
• Threading - Incorrect calculation of IdleWorkerThreadCount 

 

 

On the plus side these can be readily fixed yourself for your programs. 

 

I mean, I've been running with a patched RTL for years that fixes all the design flaws in handling of floating point control flags. 

 

At least we have access to the RTL source code and so can apply fixes easily using code hooks. 

[pyscripter 689]

4 hours ago, Tommi Prami said:

I am pretty interested in hearing what are people thinking of implementing with Managed records.

 

-Tee-

I think there is a serious issue with the timing of finalization of temp managed records which now is ASAP unlike regular records.  Please see

 

Timing of finalization of temporary managed records

 

I would love to hear what the experts think about this.  @Stefan Glienke @David Heffernan @Marco Cantu etc.  

Edited May 28, 2020 by pyscripter

[Stefan Glienke 2002]

@pyscripter Looks weird given how other managed types lifetimes are being handled - but since we don't have precise language and design specs as always we can just guess.

 

Apart from that implementing smartpointer that way is wrong anyway because they break whenever there is an explicit or implicit assignment/copy going on.

Proper smartpointer need a shared block that holds a reference counter.

 

Shared<T> in Spring is implemented with several optimizations to avoid the overhead of object creation for this purpose using the all known "interface in record" approach and I challenge everyone to come up with a robust implementation using custom managed records to beat that.

Edited May 28, 2020 by Stefan Glienke

[Anders Melander 1783]

1 hour ago, pyscripter said:

I think there is a serious issue with the timing of finalization of temp managed records which now is ASAP unlike regular records. 

I actually prefer that finalization style over the way interfaces etc. are finalized. At least this is by design while the old style seems more like a side effect of an implementation detail.

 

I've always had a problem with this:

var
  Foo: IUnknown;
begin
  Foo := TInterfacedObject.Create as IUnknown;
  ...some code here...
  Foo := nil; // This doesn't destroy the object
  ...more code here...
end; // but this does

I've not tried it but from the comment by Chee Wee Chua on RSP-28520 is seems I could solve it with a managed record inside a local block.

[pyscripter 689]

17 minutes ago, Anders Melander said:

I actually prefer that finalization style over the way interfaces etc. are finalized. At least this is by design while the old style seems more like a side effect of an implementation detail.

 

I've always had a problem with this:

var
  Foo: IUnknown;
begin
  Foo := TInterfacedObject.Create as IUnknown;
  ...some code here...
  Foo := nil; // This doesn't destroy the object
  ...more code here...
end; // but this does

I've not tried it but from the comment by Chee Wee Chua on RSP-28520 is seems I could solve it with a managed record inside a local block.

Your example is not correct:

 

Try:

program Scope;
{$APPTYPE CONSOLE}

uses
  System.SysUtils,
  System.Classes;

type
  TTestScope = class(TInterfacedObject)
     destructor Destroy; override;
  end;

{ TTestScope }
destructor TTestScope.Destroy;
begin
  WriteLn('Gone');
  inherited;
end;

procedure Test;
var
  Foo: IUnknown;
begin
  Foo := TTestScope.Create as IUnknown;
  WriteLn('Do stuff');
  Foo := nil; // This does destroy the object
  WriteLn('Do more stuff');
end;

begin
  Test();
  ReadLn;
end.

Output:

 

Do stuff
Gone
Do more stuff

 

 

Edited May 28, 2020 by pyscripter

[pyscripter 689]

@Anders Melander The general rule that has always applied is that managed types (strings, interfaces, records with managed fields, dynamic arrays, etc,) are finalized at the end of the scope in which they are introduced.  This included temp variables.

The newly introduced managed records breaks this rule.  Whether you like it better or not, it is inconsistent.

 

And a local block does not solve the problem.  The temp managed record will still self-destruct at the end of the statement it is used. 

Edited May 28, 2020 by pyscripter

[Anders Melander 1783]

6 hours ago, pyscripter said:

Your example is not correct

You're right. I tried to simplify the problem and simplified it so much that the problem didn't occur.

 

The actual problem I'm having is with factory methods that returns an interface; A common example is an implementation of the memento pattern that changes the current cursor and restores it when the memento is destroyed:

begin
  SaveCursor(crHourGlass);
  ...do stuff...
end; // Cursor is restored here

Works great, but occasionally I want to restore the cursor before the end of the block so I do like this:

var
  SavedCursor: ICursorRecall;
begin
  SavedCursor := SaveCursor(crHourGlass);
  ...do stuff...
  SaveCursor := nil;
  ...do more stuff...
end; // cursor is still restored here

It's been my understanding that this was caused by a hidden temporary variable that is created by the compiler when you call a function returning an interface, and I've just learned to live with it. The strange thing is that it doesn't always behave like this. Simple cases, like the example above, probably works as expected (i.e. nilling the reference destroys the object), but in more complex code the object is only destroyed when the reference goes out of scope.

[Guest]

Hide feeds!

 

This was a button in Tokyo and i had to press it each time i started the IDE (and restart and restart).

In Sydney it is a link in the bottom and the click on that link is persisted!

 

Neat-o.

 

/D

[mvanrijnen 123]

I like the record initialziation/finalization, but i was working on something i can use it for, and got this:

 

procedure THSNullableBooleanSerializer.Serialize(const AValue; const AWriter: IgoBsonBaseWriter);
var
  jsonvalue : THSNullableBoolean absolute AValue;
begin
    // do stuff
end;

[dcc32 Error] HESI.NullableTypes.pas(158): E2634 Declaring a managed record with absolute directive is not allowed

 

 

hmz, to bad ....