Add additional ByteBufferTest tests - #982
Open
jonpryor wants to merge 1 commit into
Open
Conversation
Context: https://github.com/dotnet/roslyn/issues/80952 Context: unoplatform/uno#24036 Context: 608fac6 / jbevain#958 I've been trying to get a project building for iOS+Native AOT, which invariably would error out in `ilc`: Error: Sequence point value is out of range. System.BadImageFormatException: Sequence point value is out of range. at System.Reflection.Throw.SequencePointValueOutOfRange() in /_/src/runtime/src/libraries/System.Reflection.Metadata/src/System/Reflection/Throw.cs:line 239 at System.Reflection.Metadata.SequencePointCollection.Enumerator.MoveNext() in /_/src/runtime/src/libraries/System.Reflection.Metadata/src/System/Reflection/Metadata/PortablePdb/SequencePointCollection.cs:line 86 at Internal.TypeSystem.Ecma.PortablePdbSymbolReader.<GetSequencePointsForMethod>d__10.MoveNext() in /_/src/runtime/src/coreclr/tools/Common/TypeSystem/Ecma/SymbolReader/PortablePdbSymbolReader.cs:line 147 at ILCompiler.Logging.MessageOrigin..ctor(MethodIL, Int32) in /_/src/runtime/src/coreclr/tools/aot/ILCompiler.Compiler/Compiler/Logging/MessageOrigin.cs:line 65 After quite a bit of investigation, determined that Cecil was apparent cause of the crash, because multiple "involved parties" did *not* have the fix present in commit 608fac6, resulting in a "failure cascade": 1. `<Csc/>` task produces `.pdb` file 2. Uno `<EmbeddedResourceInjectorTask_v0/>` task uses Cecil to update (1), introduces "Invalid compressed integer". 3. `illink` -- which has its own copy of Cecil, from the dotnet/cecil fork -- consumes (2), turning "Invalid compressed integer" values into *different* values. 4. `ilc` processes `pdb` produced from (3), fails with `BadImageFormatException: Sequence point value is out of range`. A "full stack" fix involves updating Uno to use Mono.Cecil 0.11.6, which contains 608fac6, *and also* updating `illink` to also contain the fix from 608fac6. (PR does not yet exist.) However, as part of this investigation, Claude cooked up a patch to dotnet/cecil along with unit test updates to ByteBufferTests. Of interest is that the suggested Claude fix for dotnet/cecil also causes the original ByteBufferTests test from 608fac6 to pass (yay), but not all of the new tests. Meanwhile, jbevain/cecil *does* pass all of the suggested new tests. Record the new tests for posterity, and to reduce future drift between jbevain/cecil and dotnet/cecil.
jonpryor
added a commit
to jonpryor/cecil
that referenced
this pull request
Aug 11, 2026
Context: https://github.com/dotnet/roslyn/issues/80952 Context: unoplatform/uno#24036 Context: jbevain@608fac6 Context: jbevain#982 I've been trying to get a project building for iOS+Native AOT, which invariably would error out in `ilc`: Error: Sequence point value is out of range. System.BadImageFormatException: Sequence point value is out of range. at System.Reflection.Throw.SequencePointValueOutOfRange() in /_/src/runtime/src/libraries/System.Reflection.Metadata/src/System/Reflection/Throw.cs:line 239 at System.Reflection.Metadata.SequencePointCollection.Enumerator.MoveNext() in /_/src/runtime/src/libraries/System.Reflection.Metadata/src/System/Reflection/Metadata/PortablePdb/SequencePointCollection.cs:line 86 at Internal.TypeSystem.Ecma.PortablePdbSymbolReader.<GetSequencePointsForMethod>d__10.MoveNext() in /_/src/runtime/src/coreclr/tools/Common/TypeSystem/Ecma/SymbolReader/PortablePdbSymbolReader.cs:line 147 at ILCompiler.Logging.MessageOrigin..ctor(MethodIL, Int32) in /_/src/runtime/src/coreclr/tools/aot/ILCompiler.Compiler/Compiler/Logging/MessageOrigin.cs:line 65 After quite a bit of investigation, determined that Cecil was apparent cause of the crash, because multiple "involved parties" did *not* have the fix present in commit jbevain/cecil@608fac6f, resulting in a "failure cascade": 1. `<Csc/>` task produces `.pdb` file 2. Uno `<EmbeddedResourceInjectorTask_v0/>` task uses Cecil to update (1), introduces "Invalid compressed integer". 3. `illink` -- which has its own copy of Cecil, from the dotnet/cecil fork -- consumes (2), turning "Invalid compressed integer" values into *different* values. 4. `ilc` processes `pdb` produced from (3), fails with `BadImageFormatException: Sequence point value is out of range`. A "full stack" fix involves updating Uno to use Mono.Cecil 0.11.6, which contains jbevain/cecil@608fac6f, *and also* updating `illink` to also contain the fix from jbevain/cecil@608fac6f. However, as part of this investigation, Claude cooked up a patch to dotnet/cecil along with unit test updates to ByteBufferTests. Of interest is that the suggested Claude fix for dotnet/cecil also causes the original ByteBufferTests test from jbevain/cecil@608fac6f to pass (yay), but not all of the new tests. Meanwhile, jbevain/cecil *does* pass all of the suggested new tests. Record the new tests for posterity, and to reduce future drift between jbevain/cecil and dotnet/cecil.
This was referenced Aug 11, 2026
jonpryor
added a commit
to jonpryor/cecil
that referenced
this pull request
Aug 12, 2026
Context: https://github.com/dotnet/roslyn/issues/80952 Context: unoplatform/uno#24036 Context: jbevain@608fac6 Context: jbevain#982 I've been trying to get a project building for iOS+Native AOT, which invariably would error out in `ilc`: Error: Sequence point value is out of range. System.BadImageFormatException: Sequence point value is out of range. at System.Reflection.Throw.SequencePointValueOutOfRange() in /_/src/runtime/src/libraries/System.Reflection.Metadata/src/System/Reflection/Throw.cs:line 239 at System.Reflection.Metadata.SequencePointCollection.Enumerator.MoveNext() in /_/src/runtime/src/libraries/System.Reflection.Metadata/src/System/Reflection/Metadata/PortablePdb/SequencePointCollection.cs:line 86 at Internal.TypeSystem.Ecma.PortablePdbSymbolReader.<GetSequencePointsForMethod>d__10.MoveNext() in /_/src/runtime/src/coreclr/tools/Common/TypeSystem/Ecma/SymbolReader/PortablePdbSymbolReader.cs:line 147 at ILCompiler.Logging.MessageOrigin..ctor(MethodIL, Int32) in /_/src/runtime/src/coreclr/tools/aot/ILCompiler.Compiler/Compiler/Logging/MessageOrigin.cs:line 65 After quite a bit of investigation, determined that Cecil was apparent cause of the crash, because multiple "involved parties" did *not* have the fix present in commit jbevain/cecil@608fac6f, resulting in a "failure cascade": 1. `<Csc/>` task produces `.pdb` file 2. Uno `<EmbeddedResourceInjectorTask_v0/>` task uses Cecil to update (1), introduces "Invalid compressed integer". 3. `illink` -- which has its own copy of Cecil, from the dotnet/cecil fork -- consumes (2), turning "Invalid compressed integer" values into *different* values. 4. `ilc` processes `pdb` produced from (3), fails with `BadImageFormatException: Sequence point value is out of range`. A "full stack" fix involves updating Uno to use Mono.Cecil 0.11.6, which contains jbevain/cecil@608fac6f, *and also* updating `illink` to also contain the fix from jbevain/cecil@608fac6f. However, as part of this investigation, Claude cooked up a patch to dotnet/cecil along with unit test updates to ByteBufferTests. Of interest is that the suggested Claude fix for dotnet/cecil also causes the original ByteBufferTests test from jbevain/cecil@608fac6f to pass (yay), but not all of the new tests. Meanwhile, jbevain/cecil *does* pass all of the suggested new tests. Record the new tests for posterity, and to reduce future drift between jbevain/cecil and dotnet/cecil.
sbomer
pushed a commit
to dotnet/cecil
that referenced
this pull request
Aug 12, 2026
Context: https://github.com/dotnet/roslyn/issues/80952 Context: unoplatform/uno#24036 Context: jbevain@608fac6 Context: jbevain#982 I've been trying to get a project building for iOS+Native AOT, which invariably would error out in `ilc`: Error: Sequence point value is out of range. System.BadImageFormatException: Sequence point value is out of range. at System.Reflection.Throw.SequencePointValueOutOfRange() in /_/src/runtime/src/libraries/System.Reflection.Metadata/src/System/Reflection/Throw.cs:line 239 at System.Reflection.Metadata.SequencePointCollection.Enumerator.MoveNext() in /_/src/runtime/src/libraries/System.Reflection.Metadata/src/System/Reflection/Metadata/PortablePdb/SequencePointCollection.cs:line 86 at Internal.TypeSystem.Ecma.PortablePdbSymbolReader.<GetSequencePointsForMethod>d__10.MoveNext() in /_/src/runtime/src/coreclr/tools/Common/TypeSystem/Ecma/SymbolReader/PortablePdbSymbolReader.cs:line 147 at ILCompiler.Logging.MessageOrigin..ctor(MethodIL, Int32) in /_/src/runtime/src/coreclr/tools/aot/ILCompiler.Compiler/Compiler/Logging/MessageOrigin.cs:line 65 After quite a bit of investigation, determined that Cecil was apparent cause of the crash, because multiple "involved parties" did *not* have the fix present in commit jbevain/cecil@608fac6f, resulting in a "failure cascade": 1. `<Csc/>` task produces `.pdb` file 2. Uno `<EmbeddedResourceInjectorTask_v0/>` task uses Cecil to update (1), introduces "Invalid compressed integer". 3. `illink` -- which has its own copy of Cecil, from the dotnet/cecil fork -- consumes (2), turning "Invalid compressed integer" values into *different* values. 4. `ilc` processes `pdb` produced from (3), fails with `BadImageFormatException: Sequence point value is out of range`. A "full stack" fix involves updating Uno to use Mono.Cecil 0.11.6, which contains jbevain/cecil@608fac6f, *and also* updating `illink` to also contain the fix from jbevain/cecil@608fac6f. However, as part of this investigation, Claude cooked up a patch to dotnet/cecil along with unit test updates to ByteBufferTests. Of interest is that the suggested Claude fix for dotnet/cecil also causes the original ByteBufferTests test from jbevain/cecil@608fac6f to pass (yay), but not all of the new tests. Meanwhile, jbevain/cecil *does* pass all of the suggested new tests. Record the new tests for posterity, and to reduce future drift between jbevain/cecil and dotnet/cecil.
jonpryor
added a commit
to unoplatform/uno
that referenced
this pull request
Aug 13, 2026
Fixes: #24036 Context: #23966 Context: dotnet/macios#26428 Context: dotnet/cecil#504 Context: jbevain/cecil#982 Context: jbevain/cecil@0.11.4...0.11.6 Context: jbevain/cecil@0.10.1...0.11.6 In The Beginning™ was a simple idea: why don't we have an iOS+Native AOT test stage? Thus began #23966, which promptly fell over, caught on fire, and sank into the swamp. There are two issues with #23966: 1. CI was unhappy; it errors out with …/Xamarin.Shared.targets(2073,3): error : No valid iOS code signing keys found in keychain. You need to request a codesigning certificate from https://developer.apple.com. Not sure what's happening there. It doesn't happen locally… 2. It couldn't build locally either! The initial attempt to build locally would fail in `ilc`: # note: using changes from #23966 % TFM=net11.0-ios \ NAOT=1 \ BUILD_SOURCESDIRECTORY=`pwd` \ BUILD_ARTIFACTSTAGINGDIRECTORY=`pwd`/artifacts \ sh build/test-scripts/skia-ios-uitest-build.sh … …/unoplatform/uno/src/SourceGenerators/Uno.UI.Tasks/Content/Uno.UI.Tasks.targets(487,3): warning Failed to resolve assembly `JetBrains.Annotations, Version=4242.42.42.42, Culture=neutral, PublicKeyToken=1010a0d8d6380325`: Failed to resolve assembly: 'JetBrains.Annotations, Version=4242.42.42.42, Culture=neutral, PublicKeyToken=1010a0d8d6380325' EXEC : error Sequence point value is out of range. $HOME/.nuget/packages/microsoft.dotnet.ilcompiler/11.0.0-preview.6.26359.118/build/Microsoft.NETCore.Native.targets(359,5): error MSB3073: The command ""$HOME/.nuget/packages/runtime.osx-arm64.microsoft.dotnet.ilcompiler/11.0.0-preview.6.26359.118/tools/ilc" @"obj/Release/net11.0-ios/ios-arm64/native/SamplesApp.ilc.rsp"" exited with code 1. Thus began a fair bit of investigation, some of which is mentioned in #24036 and dotnet/cecil#504. What happened was a "failure cascade": 1. The `<Csc/>` task produced a `.pdb` file. 2. Uno's `<EmbeddedResourceInjectorTask_v0/>` task would use Cecil to process the assembly and `.pdb` file from (1). Uno was using, at most, Mono.Cecil 0.11.4, which had a bug around the encoding of "compressed" integers within `.pdb` files. (This is later fixed in jbevain/cecil#982, and released as part of Mono.Cecil 0.11.6.) The updated `.pdb` file would contain "Invalid compressed integer" values. 3. Later, `illink` would run, which has it's own separate copy of Mono.Cecil -- from dotnet/cecil -- which would *also* update the `.pdb` files from (2). When `illink` ran, any "Invalid compressed integer" values would be further corrupted. 4. Later still, when `ilc` runs, it would process the `.pdb` files produced by (3), and at this point instead of "silently corrupting" data, it would error out entirely: Error: Sequence point value is out of range. System.BadImageFormatException: Sequence point value is out of range. at System.Reflection.Throw.SequencePointValueOutOfRange() in /_/src/runtime/src/libraries/System.Reflection.Metadata/src/System/Reflection/Throw.cs:line 239 at System.Reflection.Metadata.SequencePointCollection.Enumerator.MoveNext() in /_/src/runtime/src/libraries/System.Reflection.Metadata/src/System/Reflection/Metadata/PortablePdb/SequencePointCollection.cs:line 86 at Internal.TypeSystem.Ecma.PortablePdbSymbolReader.<GetSequencePointsForMethod>d__10.MoveNext() in /_/src/runtime/src/coreclr/tools/Common/TypeSystem/Ecma/SymbolReader/PortablePdbSymbolReader.cs:line 147 at ILCompiler.Logging.MessageOrigin..ctor(MethodIL, Int32) in /_/src/runtime/src/coreclr/tools/aot/ILCompiler.Compiler/Compiler/Logging/MessageOrigin.cs:line 65 (Failure cascades are fun!) *Part* of the fix here is to update Uno to use Mono.Cecil 0.11.6, which contains the fix for reading and writing "compressed" integers in `.pdb` files. That is done here. Updating Cecil will *not* fix #23966, but is a partial prerequisite. A full fix will require dotnet/cecil#504 to be merged and released, (currently merged), `illink` updated to use dotnet/cecil#504, and *then* there's dotnet/macios#26428, in which `clang++` crashes when trying to link everything even if we get past the whole `.pdb` mess (which reportedly is fixed in the not-yet-released .NET 11 RC1). (Aside: the `.pdb` mess can be avoided by NOT USING `.pdb` FILES, e.g. buil with `-p:DebugType=none`. However, who wants to do that?) …which still leaves the original "how do we get this building on CI" question… Baby steps!
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Context: dotnet/maui#37782
Context: unoplatform/uno#24036
Context: 608fac6 / dotnet/roslyn#958
I've been trying to get a project building for iOS+Native AOT, which invariably would error out in
ilc:After quite a bit of investigation, determined that Cecil was apparent cause of the crash, because multiple "involved parties" did not have the fix present in commit 608fac6, resulting in a "failure cascade":
<Csc/>task produces.pdbfileUno
<EmbeddedResourceInjectorTask_v0/>task uses Cecil to update (1), introduces "Invalid compressed integer".illink-- which has its own copy of Cecil, from the dotnet/cecil fork -- consumes (2), turning "Invalid compressed integer" values into different values.ilcprocessespdbproduced from (3), fails withBadImageFormatException: Sequence point value is out of range.A "full stack" fix involves updating Uno to use Mono.Cecil 0.11.6, which contains 608fac6, and also updating
illinkto also contain the fix from 608fac6. (PR does not yet exist.)However, as part of this investigation, Claude cooked up a patch to dotnet/cecil along with unit test updates to ByteBufferTests. Of interest is that the suggested Claude fix for dotnet/cecil also causes the original ByteBufferTests test from 608fac6 to pass (yay), but not all of the new tests. Meanwhile, jbevain/cecil does pass all of the suggested new tests.
Record the new tests for posterity, and to reduce future drift between jbevain/cecil and dotnet/cecil.