FredS

54 minutes ago, aehimself said:

Application.ProcessMessages call in a loop

Start with this, I pulled that out of more complex code but haven't tested it:

//MMWIN:MEMBERSCOPY
unit _MM_Copy_Buffer_;

interface

type
  TThreadHelper = class helper for TThread
  public
    /// <summary>
    ///   Will handle all Wait Key, Mouse and Message Inputs, also calls Synchronize as needed.
    /// </summary>
    function StartAsync: TWaitResult;
  end;


implementation

function TThreadHelper.StartAsync: TWaitResult;
var
  LHandles: TArray<THandle>;
  Rsl : Cardinal;
const cObjects = 2;
begin
  Assert(TThread.CurrentThread.ThreadID = MainThreadID, 'Must be called from Main Thread');
  Self.FreeOnTerminate := False;
  SetLength(LHandles, cObjects);
  LHandles[0] := Self.Handle;
  LHandles[1] := SyncEvent;
  self.Start();
  try
    repeat
      Rsl := MsgWaitForMultipleObjects(cObjects, LHandles[0], False, INFINITE, QS_ALLINPUT);
      case Rsl of
        WAIT_FAILED: Exit(TWaitResult.wrError);
        WAIT_TIMEOUT: Exit(TWaitResult.wrTimeOut);
        WAIT_ABANDONED_0 .. WAIT_ABANDONED_0 + cObjects: Exit(TWaitResult.wrAbandoned);
        WAIT_OBJECT_0 : Result := TWaitResult.wrSignaled;
        WAIT_OBJECT_0 + 1: CheckSynchronize(0);
        WAIT_OBJECT_0 + cObjects:  Application.ProcessMessages;
      end;
    until (Rsl = 0);
  finally
    self.Free;
  end;
end;


end.

25 minutes ago, aehimself said:

guess there's no easy solution

Either write a method which uses MsgWaitForMultipleObjects and also calls synchronize `WaitAsync` or something like that, or just use a short timeout for `TTask.Wait` and call `Application.ProcessMessages`  while not terminated..

@Mahdi Safsafi Might be simpler to just redirect System.GetMemory instead of the memory manager's GetMem which is called by other methods doing a reOutOfMemory check themselves.

7 hours ago, Lars Fosdal said:

internal laundry basket, instead of the external one

Unfortunately this means that in the next delivery no one will tell me if my shirt, a critical part of the outfit, has changed functionality 🙂

11 minutes ago, Darian Miller said:

Discount offered to switch to Delphi.

That should be doable by August 1st.. /s

1 hour ago, Quarks said:

network/web app monitoring

I believe that falls under the 'new feature' category 🙂

Can connect, but have a valid cookie.

I've had issues before with the login server they use/share..

3 hours ago, Rollo62 said:

If any double on thi

One has to appreciate that the DOS installation of 'Patch 2'  truly 'empowers application developers' 🙂

All that's missing is deployment via floppy disks..

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..

6 hours ago, mvanrijnen said:

They really should sit on their customers seat for a while

Developing Delphi with Delphi was a reasonably good guarantee of that..

26 minutes ago, Darian Miller said:

I think there is a little more buzz today

Could be, certainly there was a window of opportunity with the Community Edition.. however that dimmed quickly as bug reports spiked and waited.. still waiting..

17 minutes ago, timfrost said:

one line incomprehensible Regex

Pretty simple answer on SO `{.*?}`.. looks good to me unless Lazy and Greedy become a Feature: https://quality.embarcadero.com/browse/RSP-15402

4 hours ago, Lars Fosdal said:

my subscription renewal increases the chance

Yet a good chunk of users with subscriptions are sticking with Berlin or even earlier..  certainly since 2015 debugging and related issues are worse/dangerous.. just a few days ago I saw `ReportMemoryLeaksOnShutdown` fails in 10.4 with Patch.. surprise.. never mind the ongoing saga of users unable to move their paid versions to new hardware..

1 hour ago, Mahdi Safsafi said:

define FIX_MADEXCEPT

You're like Superman.. faster than a speeding bullet 🙂

Thanks. Perfect!

35 minutes ago, Mahdi Safsafi said:

a patch to fix the leak

Thanks, Works for me..

Unfortunately when I turn on MadExcept 'Instantly Crash on Over/Under-Run' I still get exceptions:

DUnitX - [DUnitX.TestDFX.exe] - Starting Tests.

.E.E.E.E.....E.E.E.E..

Tests Found   : 11
Tests Ignored : 0
Tests Passed  : 3
Tests Leaked  : 0
Tests Failed  : 0
Tests Errored : 8

Tests With Errors

  uTest.TDDetours.Test1
  Message: Access violation at address 0000000067104A2F in module 'madExcept64.dll'. Read of address 0000000000000000

  uTest.TDDetours.Test2
  Message: Access violation at address 0000000067104A2F in module 'madExcept64.dll'. Read of address 0000000000000000

  uTest.TDDetours.Test3
  Message: Access violation at address 0000000067104A2F in module 'madExcept64.dll'. Read of address 0000000000000000

  uTest.TDDetours.Test4
  Message: Access violation at address 0000000067104A2F in module 'madExcept64.dll'. Read of address 0000000000000000

  DFX.Win.Hooks.DUnitX.Tests.THooksTests.TestHookPublicMethod
  Message: Access violation at address 0000000067104A2F in module 'madExcept64.dll'. Read of address 0000000000000000

  DFX.Win.Hooks.DUnitX.Tests.THooksTests.TestHookPublicVirtualMethod.UseRtti
  Message: Access violation at address 0000000067104A2F in module 'madExcept64.dll'. Read of address 0000000000000000

  DFX.Win.Hooks.DUnitX.Tests.THooksTests.TestHookPublicVirtualMethod.UseMockup
  Message: Access violation at address 0000000067104A2F in module 'madExcept64.dll'. Read of address 0000000000000000

  DFX.Win.Hooks.DUnitX.Tests.THooksTests.TestHookPrivateVirtualMethod
  Message: Access violation at address 0000000067104A2F in module 'madExcept64.dll'. Read of address 0000000000000000

Hi,

After implementing v2.2 and executing my own DUnitX tests I get a Memory Leak in Berlin.

I can reproduce this with Sydney and MadExcept, oddly no Memory Leak by simply setting `ReportMemoryLeaksOnShutdown := True` in Sydney..
In addition, turning on MadExcept Leak Test causes exceptions..

These changes where made to uTest to duplicate the issue:

var
  TDDetoursTrampolineAdd: TAdd = nil;
  
procedure TDDetours.Setup;
var LTransaction : THandle;
begin
  ReportMemoryLeaksOnShutdown := True;
  LTransaction := DDetours.BeginTransaction;
  TDDetoursTrampolineAdd := InterceptCreate(@Add, @InterceptAdd);
  DDetours.EndTransaction(LTransaction);
end;

procedure TDDetours.TearDown;
var LTransaction : THandle;
begin
  LTransaction := DDetours.BeginTransaction;
  InterceptRemove(@TDDetoursTrampolineAdd);
  DDetours.EndTransaction(LTransaction);
end;

allocation number: 11072
program up time: 1.07 s
type: GetMem
address: $2c9f820
size: 24
access rights: read/write

main thread ($b74):
671a5da7 madExcept32.dll madExceptDbg       1743 GetMemCallback
0041463e Test.exe        System            39806 GetMemory
006e4ce1 Test.exe        DDetours           2595 BeginTransaction
006ab856 Test.exe        DUnitX.TestRunner       TDUnitXTestRunner.Loggers_TeardownTest
006e5836 Test.exe        uTest               100 TDDetours.TearDown
006ab0ca Test.exe        DUnitX.TestRunner       TDUnitXTestRunner.ExecuteTestTearDown
006aadc7 Test.exe        DUnitX.TestRunner       TDUnitXTestRunner.ExecuteTests
006aa37e Test.exe        DUnitX.TestRunner       TDUnitXTestRunner.ExecuteFixtures
006aa3ab Test.exe        DUnitX.TestRunner       TDUnitXTestRunner.ExecuteFixtures
006aa010 Test.exe        DUnitX.TestRunner       TDUnitXTestRunner.Execute
006ee2ca Test.exe        Test                 48 initialization
7507343b kernel32.dll                            BaseThreadInitThunk

memory dump: 
02c9f820  34 0c 00 00 30 0e 00 00 - 9b 9e 73 f3 50 f8 f2 e9  4...0.....s.P...
02c9f830  9a 53 61 bd 93 e4 94 be  

1 hour ago, t2000 said:

see 1st post

I did, but I don't actually know what TWaitFor does..

I can tell you that the code you posted has a chance of being stuck not synchronizing.

For a quick test replace that with `CheckSynchronize(10)`.

Optimal would be a MsgWaitForMultipleObjects waiting on Thread or Process Termination and the SyncEvent while handling messages.

23 minutes ago, t2000 said:

WaitMessage

What does that do?

21 minutes ago, Anders Melander said:

I haven't encountered any issues that I can pin on Windows 7

@Anders Melander

Yup, certainly no issues not already reported except maybe that the History list box is all black and can't be used.. since I haven't seen this reported I'm not sure if its W7 or RDP..
All in all RDP'ing is better out of the box in 10.4, 10.3 was useless without a fix from Andreas..

10 hours ago, dummzeuch said:

I can't meaningfully compare them

It's faster especially when you are only enumerating folders..

To do a real test one would have to access remote ADMIN$ shares, something like a 2012 server or newer with a huge directory tree..
Run a few of those on Azure and run them simultaneously (multi-threaded).. but not in 10.4 .. being able to retrieve the data in large chunks makes enough of a difference..

8 hours ago, Scott said:

Seems like Embarcadero need to increase their Beta testing pool

That would work if bugs found where actually fixed instead of disappearing into the internal pool of darkness..

On 5/29/2020 at 12:34 PM, Stéphane Wierzbicki said:

Waiting now for Andreas to fix this

After a couple of clicks on components the structure window no longer follows your selection.
Right about then I realized that the structure window 'Find' control comes from DDevExtensions (@AndyHTech).. 🙂

2 hours ago, dummzeuch said:

several times already

I've seen it as a hint 'LSP has stopped working' or something similar.

Only in one unit so far.. apparently new declarations in Winapi.Windows for some structures duplicated JwaWindows and that did it.

ErrorInsight fails when you alter something and Codeinsight breaks after a couple of debug sessions..
 

But the most annoying is the keyboard.. maybe its because I'm trial running on an Unsupported OS (Windows 7) 🙂

1 hour ago, timfrost said:

search a Samba network drive

This works fine, and I also use a home rolled method using FindFirstFileEx.

As for multiple threads, I did that test a while back and it didn't show any better time.

12 minutes ago, Kas Ob. said:

after a given amount of time you stop being a current paying customer, you lose access or download options.

That part sounds fair enough to me..

More troubling is that the New Embarcadero Licenses and Download Portal still has no method of handling license bumps as outlined by the post above yours.
As it sits now I can be responsible and burn an ISO or make backups but I can't find a legal way to move my installation..