Rollo62

I use such "invisible" form in one of my projects. So that the styles are still be able to edit there, sometimes to make custom edits. I uses that form for editing, and removed all unneeded stuff later, but keep that StyleBook on that form for the whole application. I'm not happy with that either, but it does its job, and its more convenient than to have a separate form elsewhere, just for designer purposes. In a DataModule the stylebook designer didn't work as it should, at least you cannot edit any custom styles, or is there any workaround to it ? I doubt it, because DataModule is not visual.

Dear Remy, thanks for the info, there is a nice tutorial howto do that in XCode. https://forum.unity.com/threads/xcode-storyboard-option-for-splash-screens-launch-screens-in-ios-build.811131/ Have not checked it, but I hope that works for me too, no idea yeat what else might this change affect. Will there be any special requirements to the images itself (size, type, etc.) ?

Call it: .IsAdministrator and all is clear

Usually such mess can start if you have cyclic unit definitions. I would try to cleanup and rearrange used units, to maybe find the possible root cause.

I didnt say And64 makes no sense, I just pointed out that this problably helps to solve some issues.

I think Android64 doesn't even start on Android32 OS, and the selection of libraries can be done in the deployment.

Maybe this will be there in Version 10.4 ...

Thanks god the new managements didn't kill all old references in code and IDE as first step, causing all kinds of weired issues.

Have a look into the KastriFree project from Dave Nottage.

No time to do that at the moment, but my current setting was produced from such clean Windows image. Only I had some updates now, and its not that clean machine anymore. Maybe I will check later, I have no 3D applications at the moment, and the IDE is running well. Better I wait until Rx10.4, then I have a good reason to start with a clean installation again.

Thanks for testing so deeply ☝️ Nothing, I love inline variables too. I just would wait until its really stable for production, maybe I look at it in Rx10.4.

By the way, I just found out that my VmWare also crashes when VmWare+IDE is running, and when I start then PAServer via PAServerManager on the same host. But its a different crash type, since I can close and restart VmWare normally, but maybe this is related. When I start PAServer first, and then run VmWare+IDE, all is OK. So its a matter of startup sequence here. P.S. I just see there is a new VmWare version 11.5.2 available, I will check that out later. https://docs.vmware.com/en/VMware-Fusion/11.5.0/rn/VMware-Fusion-1152-Release-Notes.html

Thanks, yes I have seen that link in the bottom. I referred to it here Does this leads to the same page in your system ? But the hints doesn'T help, as I wrote, I cannot find any Vulkan.xxx libraries under Windows/System32, etc. I think that VmWare simply is not ready yet for the full Vulkan support under Windows 10 (my assumnption).

Short update: I've got proposals from VmWare support: 1. ReInstall VmWare tools 2. Disable 3D acceleration 1. I've tried, no success. That was forseeable for me, since I update VmTools usually always after updates (und usually I don't forget that). 2. Disable 3D acceleration helps of coarse, but switched off 3D acceleration. My 3D apps work again. Not 100% what 3D acceleration means under VmWare Fusion, but I assume this switches to new Metal driver. If disabled, I assume this falls back to OpenGL driver. https://docs.vmware.com/en/VMware-Fusion/11/com.vmware.fusion.using.doc/GUID-C0E9FDAC-BC40-4A6B-8940-013597CA5E5B.html https://docs.vmware.com/en/VMware-Fusion/11.5.0/rn/VMware-Fusion-1150-Release-Notes.html https://communities.vmware.com/thread/581677 https://communities.vmware.com/message/2562929#2562929 Anyway, maybe I have to wait until Metal drivers are more stable, no big issue for me at the moment. Of coarse I don't want to see any sudden, fatal crashes anymore.

@Fr0sT.Brutal Yes, did that before. Tends to get very ugly and messy with ios and Android soon. Thats why I changed my method.

I try to separate all code into platform separated untits meanwhile, even if this can get more work. But the IFDEFs tend to get messy very soon, even in small units. Even if the units are getting very small and simple, I found for myself that its very worth it to get a clean structure. unit MyUnit type TMyBaseOrAbstractClass = class or better an interface end; uses {$IF DEFINED( ANDROID )} MyUnit.Impl.Android, {$ELSEIF DEFINED( IOS )} MyUnit.Impl.iOS, {$ELSEIF DEFINED( Macos )} MyUnit.Impl.Macos, {$ELSEIF DEFINED( MSWINDOWS )} MyUnit.Impl.Win, {$ELSEIF DEFINED( Linux )} MyUnit.Impl.Linux, {$ELSE} {$MESSAGE ERROR 'MyUnit: unkown platform' } {$ENDIF} FMX.DialogService; ... unit MyUnit.Android uses MyUnit; type TMyBaseOrAbstractClass_Impl = class( TMyBaseOrAbstractClass ) end;

Did you "upgrade" your old project from 32 to 64-Bit ? As the platform 64 was added, the automatic conversion of the IDE might fail. As vfbb proposes, check the libraries. Maybe even better, try to completely create a new project file, as 10.3.3 might put all necessary files into it.

Thanks for testing. Strange, I have the same version, but it reliably crashes on 3D acceleration. What I found is that the whole VmWare app crashes, also when opening Paint3D or some other 3D related apps. Maybe RadStudio switches to 3D hardware acceleration mode only when a project is loaded, that would make sense. It seems something wrong with my VmWare 3D drivers here, but I haven't changed anything special. I have to check with VmWare more deeply.

Is the app in the PlayStore already switched to AAB Bundles, or do you try to update APK ?

Ups, the scene doesn't look like that much. What you mean by "main scene" ? What I meant was just loading liles needed for the current view, not keeping in background. When switching from scene1 to scene2 you have maybe enough time to load, even with a little delay, but this is maybe not visible or you could cover this by some scene change splash image. If scenes move linear, you could hold and prepare the last and next view maybe in the background, while running the current view. If you have one view really needing 500MB, OK then I'm out

Still don't understand why you need to reload on-the-fly, instead of pre-load all relevant bitmap lists, and just switch them on demand.

Hi there, I've got a strange exception, only with Android 32-Bit during Debugging, other platforms including IOS work well. With iOS 64 I had no issue seen, but wouldn't count on it right now. I use a TDateTime variable, and want to compare against NULL.. Since TDateTime is a Double with 8 Bytes, a simple cast to Double( Self ) should make not any harm. Maybe there are special conditions on Android, as Extended is reduced to Double, and maybe there are some conversion side effects with Double as well ? I use a class helper for adding such functionality, which is used in a million other places too. function TDateTime_Helper.ToDouble : Double; begin Result := Double( Self ); // Here it crashes, see images enclosed, they only have 2 ARM assembler lines end; and I already extended my conversion routing to separated local variables: // global variable for storage, only reading var LUNull : TDateTime = TDateTime( 0.0 ); class function TDateTime_Helper.Create_Null : TDateTime; begin Result := LUNull; // Could the global variable cause issues in a Thread ? Buts its readonly. end; function TDateTime_Helper.IsNull : Boolean; var LSelf : Double; LNull : Double; begin LSelf := Self.ToDouble; //11.03.20 added local variables, to check Android crash LNull := Create_Null.ToDouble; if SameValue( LSelf, LNull ) then Result := True else Result := False; end; I check a variable, which is called in a thread if FLastUpdate.IsNull then // called like this begin .... end I must confess that I use above scheme in many thousands of places, also heavily within threads (even higher loded), without an error before. The caller can be debugged very reliable, and the error is very reproducable When the caller comes in first, second, third time, with zero, all is fine When the caller comes in with a real Double value, I can debug and see a valid double value in the watch (e.g. FLastUpdate = 43901.1603147338 ) Inside the conversion ToDouble it crashes When entering with Zero, it looks like the images img. 1 and 2 (see the registers changing) When entering with Date, it looks like img. 3, and immediately it crashes at img. 4 What could cause such error, it sounds a little like failure of JNI Bus, but maybe I'M totally on the wrong track ? Probably there is some genius with a good hint how to fix this nasty bug.

I just tired a brute-force approach to get this solved. I replaced this function TDateTime_Helper.ToDouble : Double; begin Result := Double( Self ); end; by that function TDateTime_Helper.ToDouble : Double; var LDbl : Double; begin Move( Self, LDbl, Sizeof( LDbl) ); Result := LDbl; end; And believe it or not, I have a crash-less app now. Any ideas or explanations ? I still would like to get back to the original approach, which should be more effective (untested).

I do create my files with BOM too, and additionally try to keep existing status when loading alien documents. That way, all parties should be happy. I try to follow these rules: File has BOM --> Keep BOM File without BOM --> Keep without BOM Create my own file --> Always use BOM

Means that it is not working at all, or that Google noted some devices are not compatible ? If the latter, you can find some good info in the PreLaunchTest page.