# Native EBuild-generated image and icon resources fail at runtime on .NET 10 — BinaryFormatter dependency

**URL:** https://talk.remobjects.com/t/native-ebuild-generated-image-and-icon-resources-fail-at-runtime-on-net-10-binaryformatter-dependency/34093
**Category:** Oxygene
**Tags:** visual-studio, oxygene
**Created:** [October 6, 2026, 12:10am UTC](https://talk.remobjects.com/t/native-ebuild-generated-image-and-icon-resources-fail-at-runtime-on-net-10-binaryformatter-dependency/34093 "2026-10-06T00:10:49Z")
**Posts on this page:** 19
**Page:** 1

<div class="post-metadata">

### Author: ![claudioh](https://talk.remobjects.com/letter_avatar_proxy/v4/letter/c/8edcca/32.png) [@claudioh](https://talk.remobjects.com/u/claudioh)
#### Post date: [October 6, 2026, 12:10am UTC](https://talk.remobjects.com/t/native-ebuild-generated-image-and-icon-resources-fail-at-runtime-on-net-10-binaryformatter-dependency/34093/1 "2026-10-06T00:10:49Z")

</div>

I would like to report a resource compatibility issue with .NET 10 that appears to be distinct from the RESX regression I previously reported.

The previous issue affected .NET Framework 4.8 after upgrading to Elements **13.0.0.3125** , producing the error “The input stream is not a valid binary format.” Your team reproduced that regression and announced a fix. Downgrading to **13.0.0.3081** restored the .NET Framework application’s behavior.

However, the following .NET 10 issue also occurs with **13.0.0.3081**.

Environment:

- Oxygene / Elements / EBuild **13.0.0.3081**
- .NET SDK **10.0.302**
- Microsoft.WindowsDesktop.App **10.0.10**
- Target: **.NET 10, Windows, x86**
- Visual Studio 2026; the reproduction was built directly with EBuild.

We created an isolated Oxygene test project with a newly created `.resx` containing:

- A string.
- A `ResXFileRef` referencing a newly generated PNG as `System.Drawing.Bitmap`.
- A `ResXFileRef` referencing an ICO file as `System.Drawing.Icon`.

These were new test resources, not resources copied from our legacy application.

The project uses the standard `EmbeddedResource` approach, as shown in the installed WinForms Core template, without external resource generation:

```auto
<TargetFramework>.NET10.0-windows</TargetFramework>
<NETCoreRuntime>Microsoft.WindowsDesktop.App</NETCoreRuntime>
<NETCoreRuntimeVersion>10.0.10</NETCoreRuntimeVersion>
<NETCoreSDKVersion>10.0.302</NETCoreSDKVersion>

```

```auto
<EmbeddedResource Include="Resources.resx">
  <LogicalName>ProvaRecursos.Resources.resources</LogicalName>
</EmbeddedResource>

```

After a successful build, the test reads each resource using `System.Resources.ResourceManager.GetObject`.

Results:

| Test | Result |
| --- | --- |
| Native EBuild compilation | Successful |
| String resource | Loaded successfully |
| Bitmap resource | Failed |
| Icon resource | Failed |

Both the bitmap and icon fail with:

```auto
System.NotSupportedException:
BinaryFormatter serialization and deserialization are disabled within this application.
See https://aka.ms/binaryformatter for more information.

```

In our actual .NET 10 application, generating resources separately through the Microsoft SDK’s `GenerateResource` task with `UsePreserializedResources="true"` works. EBuild then embeds the generated `.resources` files.

However, our goal is to use native resource processing directly from `.resx`, as we would expect for a new Oxygene .NET 10 application, without maintaining a separate resource-generation project and resource list.

Could you please clarify:

1. Is there a supported EBuild setting to generate image and icon resources that load correctly on .NET 10 without enabling BinaryFormatter?
2. Is the direct `EmbeddedResource` approach the recommended configuration for a new Oxygene WinForms application targeting .NET 10?
3. Is this behavior related to the previously acknowledged RESX regression, or does it require a separate fix?

Thank you.

---

<div class="post-metadata">

### Author: ![mh](https://talk.remobjects.com/user_avatar/talk.remobjects.com/mh/32/18050_2.png) [@mh](https://talk.remobjects.com/u/mh)
#### Post date: [October 6, 2026, 10:09am UTC](https://talk.remobjects.com/t/native-ebuild-generated-image-and-icon-resources-fail-at-runtime-on-net-10-binaryformatter-dependency/34093/2 "2026-10-06T10:09:53Z")

</div>

Reated to the previous reort, but not quite the same issue indeed. we’re investigating.

---

<div class="post-metadata">

### Author: ![mh](https://talk.remobjects.com/user_avatar/talk.remobjects.com/mh/32/18050_2.png) [@mh](https://talk.remobjects.com/u/mh)
#### Post date: [October 6, 2026, 2:05pm UTC](https://talk.remobjects.com/t/native-ebuild-generated-image-and-icon-resources-fail-at-runtime-on-net-10-binaryformatter-dependency/34093/3 "2026-10-06T14:05:31Z")

</div>

I’ve uploaded `RemObjects Elements with Water - 13.0.0.3138.exe`, can you check if both issues are addressed for you?

---

<div class="post-metadata">

### Author: ![claudioh](https://talk.remobjects.com/letter_avatar_proxy/v4/letter/c/8edcca/32.png) [@claudioh](https://talk.remobjects.com/u/claudioh)
#### Post date: [October 6, 2026, 2:28pm UTC](https://talk.remobjects.com/t/native-ebuild-generated-image-and-icon-resources-fail-at-runtime-on-net-10-binaryformatter-dependency/34093/4 "2026-10-06T14:28:39Z")

</div>

Yes, I tested Elements **13.0.0.3138** , but encountered additional compilation problems that prevented me from validating the original RESX fix at runtime.

Our previous version was **13.0.0.3081**. With 3138, the builds reported:

- **.NET Framework 4.8 Client:** 13 errors.
- **.NET 10 Server:** 4,115 errors, potentially including cascading diagnostics.

The errors affect existing Oxygene code and appear to involve the resolution of parameterless method calls without parentheses. Expressions previously evaluated as method calls now appear to be treated as method groups.

Examples from the compiler output:

```auto
Parameter 1 is "<method group>", should be "IWin32Window",
in call to XtraBaseArgs.<Write>Owner(value: IWin32Window)

```

```auto
No overloaded method "Append" with these parameters on type "StringBuilder",
best matching overload is "Append(value: UInt64): StringBuilder"

```

```auto
No overloaded method "ToString" with these parameters on type "Double",
best matching overload is "ToString(provider: IFormatProvider): String"

```

There are also similar errors involving `DateTime.ToString`, `String.Format`, and `MessageBox.Show`.

The builds were run directly through EBuild. We have not changed the affected calls or compiler settings to work around this behavior.

Could this be another regression in 3138 related to implicit invocation of parameterless methods?

---

<div class="post-metadata">

### Author: ![mh](https://talk.remobjects.com/user_avatar/talk.remobjects.com/mh/32/18050_2.png) [@mh](https://talk.remobjects.com/u/mh)
#### Post date: [October 6, 2026, 3:29pm UTC](https://talk.remobjects.com/t/native-ebuild-generated-image-and-icon-resources-fail-at-runtime-on-net-10-binaryformatter-dependency/34093/5 "2026-10-06T15:29:44Z")

</div>

> [@claudioh](#):
>
> Examples from the compiler output:
> 
> ```auto
> Parameter 1 is "<method group>", should be "IWin32Window",
> in call to XtraBaseArgs.<Write>Owner(value: IWin32Window)
> 
> ```
> 
> ```auto
> No overloaded method "Append" with these parameters on type "StringBuilder",
> best matching overload is "Append(value: UInt64): StringBuilder"
> 
> ```
> 
> ```auto
> No overloaded method "ToString" with these parameters on type "Double",
> best matching overload is "ToString(provider: IFormatProvider): String"
> 
> ```
> 
> There are also similar errors involving `DateTime.ToString`, `String.Format`, and `MessageBox.Show`.

Curious, these are unrelated; could I see the code snippets that cause these? Not much has changed compiler-side since **13.0.0.3125** , I believe, aside from the Delphi compatibility changes that landed; these sound possibly related…

---

<div class="post-metadata">

### Author: ![claudioh](https://talk.remobjects.com/letter_avatar_proxy/v4/letter/c/8edcca/32.png) [@claudioh](https://talk.remobjects.com/u/claudioh)
#### Post date: [October 6, 2026, 3:40pm UTC](https://talk.remobjects.com/t/native-ebuild-generated-image-and-icon-resources-fail-at-runtime-on-net-10-binaryformatter-dependency/34093/6 "2026-10-06T15:40:32Z")

</div>

> [@mh](#):
>
> urious, these are unrelated; could I see the code snippets that cause these? Not much has changed compiler-side since **13.0.0.3125** , I believe, aside from the Delphi compatibility changes that landed; these sound possibly related…

Here are three actual source excerpts that produce errors with **13.0.0.3138**.

Both affected projects have this setting enabled:

```auto
<DelphiCompatibility>True</DelphiCompatibility>

```

For clarity, our previous working version was **13.0.0.3081** ; we are not comparing against 3125.

**1. Parameterless `ToString` passed to `StringBuilder.Append`**

Source: [ChamadaController.pas, line 219 (line 219)](/C:/Sistemas/ProjetoPegasusOOPrism/PegasusCore/PegasusServerCore/Controllers/ChamadaController.pas:219)

```auto
var TextoToolTip := new StringBuilder;
for DR: DataRow in DTGuardiao.Rows do
  TextoToolTip.Append(DR['CODLOJA'].ToString)
               .Append(' - ')
               .Append(DR['NOMELOJA'].ToString)
               .Append(DR['COMENTARIO'].ToString)
               .Append(Environment.NewLine);

```

Compiler output:

```auto
No overloaded method "Append" with these parameters on type "StringBuilder",
best matching overload is "Append(value: UInt64): StringBuilder"

parameter 1 is "<method group>" should be "UInt64"

```

**2. Parameterless method returning `IWin32Window`**

Source: [USoftSystemINIGenerica.pas, line 8513 (line 8513)](/C:/Sistemas/ProjetoPegasusOOPrism/UnitsCompartilhadas/USoftSystemINIGenerica.pas:8513)

The method is declared as:

```auto
method get_MessageBoxOwner: System.Windows.Forms.IWin32Window;

```

The failing call is:

```auto
var args : XtraMessageBoxArgs = new XtraMessageBoxArgs();
args.Owner := get_MessageBoxOwner;

```

Compiler output:

```auto
Parameter 1 is "<method group>", should be "IWin32Window",
in call to XtraBaseArgs.<Write>Owner(value: IWin32Window)

```

**3. Parameterless `ToString` used as a numeric format string**

Source: [UCGeraRelatorios.pas, line 1105 (line 1105)](/C:/Sistemas/ProjetoPegasusOOPrism/Client/Relatorio/UCGeraRelatorios.pas:1105)

Here, `Valores` is an `array of Double`, and `DTCampos` is a `DataTable`:

```auto
Valor := DTCampos.rows[i]['TextoTotalizacao'].tostring + ' ' +
         Valores[I].ToString(DTCampos.rows[i]['Mascara'].tostring);

```

Compiler output:

```auto
No overloaded method "ToString" with these parameters on type "Double",
best matching overload is "ToString(provider: IFormatProvider): String"

parameter 1 is "<method group>" should be "IFormatProvider"

```

The common pattern appears to be a parameterless method referenced without `()`, where its return value is expected. These examples come directly from our projects; we have not yet tested whether adding parentheses or disabling Delphi compatibility changes the result.

---

<div class="post-metadata">

### Author: ![mh](https://talk.remobjects.com/user_avatar/talk.remobjects.com/mh/32/18050_2.png) [@mh](https://talk.remobjects.com/u/mh)
#### Post date: [October 6, 2026, 4:08pm UTC](https://talk.remobjects.com/t/native-ebuild-generated-image-and-icon-resources-fail-at-runtime-on-net-10-binaryformatter-dependency/34093/7 "2026-10-06T16:08:27Z")

</div>

> [@claudioh](#):
>
> Both affected projects have this setting enabled:
> 
> ```auto
> <DelphiCompatibility>True</DelphiCompatibility>
> 
> ```

Ah, okay, that explains it, I think. My colleague is having a look to see whether these are regressions or expected changes.

Does adding () to the methid call [should not be necessary, i agree] alleviate any of these trhee cases?

---

<div class="post-metadata">

### Author: ![claudioh](https://talk.remobjects.com/letter_avatar_proxy/v4/letter/c/8edcca/32.png) [@claudioh](https://talk.remobjects.com/u/claudioh)
#### Post date: [October 6, 2026, 4:26pm UTC](https://talk.remobjects.com/t/native-ebuild-generated-image-and-icon-resources-fail-at-runtime-on-net-10-binaryformatter-dependency/34093/8 "2026-10-06T16:26:11Z")

</div>

es, adding `()` resolves these three cases.

However, requiring this change would disrupt our maintenance workflow. We regularly compile older versions of our system to release fixes. Those versions still contain the existing syntax, so we would have to keep switching Oxygene versions to build them.

Preserving compatibility with this syntax would therefore be important for us.

---

<div class="post-metadata">

### Author: ![mh](https://talk.remobjects.com/user_avatar/talk.remobjects.com/mh/32/18050_2.png) [@mh](https://talk.remobjects.com/u/mh)
#### Post date: [October 6, 2026, 5:34pm UTC](https://talk.remobjects.com/t/native-ebuild-generated-image-and-icon-resources-fail-at-runtime-on-net-10-binaryformatter-dependency/34093/9 "2026-10-06T17:34:19Z")

</div>

> [@claudioh](#):
>
> However, requiring this change would disrupt our maintenance workflow. We regularly compile older versions of our system to release fixes. Those versions still contain the existing syntax, so we would have to keep switching Oxygene versions to build them.

I agree. I asked this only to help narrow it down, not to imply thats the actual solution. This is Pascal, not C. Using Delphi compatibility should not make it _less_ compatible with Pascal conventions ;).

`()` should only be required in extreme ambiguituy cases (eg calling a delegate vs. using its value)

---

<div class="post-metadata">

### Author: ![claudioh](https://talk.remobjects.com/letter_avatar_proxy/v4/letter/c/8edcca/32.png) [@claudioh](https://talk.remobjects.com/u/claudioh)
#### Post date: [October 7, 2026, 10:49am UTC](https://talk.remobjects.com/t/native-ebuild-generated-image-and-icon-resources-fail-at-runtime-on-net-10-binaryformatter-dependency/34093/10 "2026-10-07T10:49:56Z")

</div>

were you able to identify this case of ( )

---

<div class="post-metadata">

### Author: ![mh](https://talk.remobjects.com/user_avatar/talk.remobjects.com/mh/32/18050_2.png) [@mh](https://talk.remobjects.com/u/mh)
#### Post date: [October 7, 2026, 12:26pm UTC](https://talk.remobjects.com/t/native-ebuild-generated-image-and-icon-resources-fail-at-runtime-on-net-10-binaryformatter-dependency/34093/11 "2026-10-07T12:26:00Z")

</div>

Uploaded a new `RemObjects Elements - 13.0.0.3138 (B).exe` for you that has the resources fixes AND th compiler regression fixes. 🤞

No designer yet 😉

---

<div class="post-metadata">

### Author: ![claudioh](https://talk.remobjects.com/letter_avatar_proxy/v4/letter/c/8edcca/32.png) [@claudioh](https://talk.remobjects.com/u/claudioh)
#### Post date: [October 7, 2026, 1:29pm UTC](https://talk.remobjects.com/t/native-ebuild-generated-image-and-icon-resources-fail-at-runtime-on-net-10-binaryformatter-dependency/34093/12 "2026-10-07T13:29:34Z")

</div>

Hi Marc,

We tested **13.0.0.3138 (20261007-105612)**:

- **Both RESX issues passed:** images/icons load correctly in .NET Framework 4.8, and native EBuild resources load in .NET 10 without the previous `BinaryFormatter` errors. Our complete client also passed: **695 resource entries loaded successfully** , without our custom resource-generation workaround.

- **One method-call issue remained:** `ToStr` is our parameterless extension method on `Object`, returning `String`. This previously valid code failed:

- **We found an encoding regression:** Windows-1252 `.pas` files without BOM, which worked in **3081** , now produce corrupted literals:

Is the encoding change intentional? Is there a compiler/project setting to retain Windows-1252 compatibility? This matters when rebuilding older branches for maintenance fixes.

Thanks!

---

<div class="post-metadata">

### Author: ![mh](https://talk.remobjects.com/user_avatar/talk.remobjects.com/mh/32/18050_2.png) [@mh](https://talk.remobjects.com/u/mh)
#### Post date: [October 7, 2026, 3:33pm UTC](https://talk.remobjects.com/t/native-ebuild-generated-image-and-icon-resources-fail-at-runtime-on-net-10-binaryformatter-dependency/34093/13 "2026-10-07T15:33:49Z")

</div>

> [@claudioh](#):
>
> - **One method-call issue remained:** `ToStr` is our parameterless extension method on `Object`, returning `String`. This previously valid code failed:

@david.millington ^^^

---

<div class="post-metadata">

### Author: ![claudioh](https://talk.remobjects.com/letter_avatar_proxy/v4/letter/c/8edcca/32.png) [@claudioh](https://talk.remobjects.com/u/claudioh)
#### Post date: [October 8, 2026, 5:54pm UTC](https://talk.remobjects.com/t/native-ebuild-generated-image-and-icon-resources-fail-at-runtime-on-net-10-binaryformatter-dependency/34093/14 "2026-10-08T17:54:51Z")

</div>

This encoding issue is causing me quite a bit of trouble, if you can get a hotfix for it as well

---

<div class="post-metadata">

### Author: ![mh](https://talk.remobjects.com/user_avatar/talk.remobjects.com/mh/32/18050_2.png) [@mh](https://talk.remobjects.com/u/mh)
#### Post date: [October 8, 2026, 9:16pm UTC](https://talk.remobjects.com/t/native-ebuild-generated-image-and-icon-resources-fail-at-runtime-on-net-10-binaryformatter-dependency/34093/15 "2026-10-08T21:16:45Z")

</div>

We’re working on it.

---

<div class="post-metadata">

### Author: ![mh](https://talk.remobjects.com/user_avatar/talk.remobjects.com/mh/32/18050_2.png) [@mh](https://talk.remobjects.com/u/mh)
#### Post date: [October 9, 2026, 4:01pm UTC](https://talk.remobjects.com/t/native-ebuild-generated-image-and-icon-resources-fail-at-runtime-on-net-10-binaryformatter-dependency/34093/16 "2026-10-09T16:01:09Z")

</div>

`RemObjects Elements - 13.0.0.3138 (G).exe` should include this fix as well. As will today’s official build.

---

<div class="post-metadata">

### Author: ![claudioh](https://talk.remobjects.com/letter_avatar_proxy/v4/letter/c/8edcca/32.png) [@claudioh](https://talk.remobjects.com/u/claudioh)
#### Post date: [October 9, 2026, 5:51pm UTC](https://talk.remobjects.com/t/native-ebuild-generated-image-and-icon-resources-fail-at-runtime-on-net-10-binaryformatter-dependency/34093/17 "2026-10-09T17:51:56Z")

</div>

> [@mh](#):
>
> `RemObjects Elements - 13.0.0.3138 (G).exe` should include this fix as well. As will today’s official build.

I repeated the original minimal reproduction using byte-for-byte identical source files, a fresh output directory and `--rebuild`.

| Source file encoding | .NET Framework 4.8 | .NET 10 |
| --- | --- | --- |
| Windows-1252, without BOM | Incorrect characters | Incorrect characters |
| UTF-8 with BOM | Correct | Correct |
| UTF-8 without BOM | Correct | Correct |

For example, this string literal in a Windows-1252 source file:

```auto
Memória Física Instalada | Configurações do Servidor | Descrição

```

Returns the following when executing the compiled method:

```auto
Mem�ria F�sica Instalada | Configura��es do Servidor | Descri��o

```

The replacement characters are **U+FFFD in the compiled string** , so this is not merely a console or UI display issue.

The original source bytes are valid Windows-1252 and remain unchanged. The behavior still appears consistent with interpreting those bytes as UTF-8.

Could you please investigate this remaining issue in build **20261009-115312 / 21ee3c5**?

---

<div class="post-metadata">

### Author: ![mh](https://talk.remobjects.com/user_avatar/talk.remobjects.com/mh/32/18050_2.png) [@mh](https://talk.remobjects.com/u/mh)
#### Post date: [October 9, 2026, 6:00pm UTC](https://talk.remobjects.com/t/native-ebuild-generated-image-and-icon-resources-fail-at-runtime-on-net-10-binaryformatter-dependency/34093/18 "2026-10-09T18:00:28Z")

</div>

This is unrelated and is “as designed”.

Oxygene was changed to treat BOM-less source files as UTF-8 rather than “system default”. This was becoming necessary because the system default differs widely (e.g., on macOS it is UTF-8), and we kept having errors where files were just fine when built and worked on on Mac, and then building the same file on Windows would fail (best case) or have bad encodings in string literals (worse). Having a consistent default across platforms was imperative.

Three options:

- save your files with a proper BOM
- save your files as UTF-8 (w/ or w/o BOM, though I’d suggest with.)
- Set the `DefaultCodePage` setting in your project to the code page you want BOM-less files to default to.

Option two with BOM is my suggested option. Fire/Water already always saves files with a proper BOM.

---

<div class="post-metadata">

### Author: ![claudioh](https://talk.remobjects.com/letter_avatar_proxy/v4/letter/c/8edcca/32.png) [@claudioh](https://talk.remobjects.com/u/claudioh)
#### Post date: [October 9, 2026, 6:20pm UTC](https://talk.remobjects.com/t/native-ebuild-generated-image-and-icon-resources-fail-at-runtime-on-net-10-binaryformatter-dependency/34093/19 "2026-10-09T18:20:41Z")

</div>

I’m going to adapt to the new encoding method and analyze which option works best for my situation.