decimal has no IL literal: legacy csc compiles 0m to a Decimal.Zero
field load, which already decompiled to the literal, but Roslyn
compiles it to a zero-initialization, which decompiled to
default(decimal). The two forms are bit-identical for decimal, so the
literal is no less faithful to the IL and matches what a human writes;
it also removes the per-compiler split in the CompoundAssignmentTest
fixture.
Assisted-by: Claude:claude-fable-5:Claude Code
Operator precedence made the condition read
(setting && Add) || Subtract, so 'x -= 1' was converted to 'x--' even
with IntroduceIncrementAndDecrement disabled. Only observable with the
non-default setting, which no fixture configuration exercises, so no
test accompanies the fix.
Assisted-by: Claude:claude-fable-5:Claude Code
Reviving the 2020 test-cases-fp-types fixtures (compound assignment on
float, double and decimal) exposed an asymmetry: post-increment and
post-decrement on float/double round-tripped as x++/x--, but the pre
forms came back as x += 1f because the increment detection in
PrettifyAssignments only accepted integer constants. C# defines ++/--
on floating-point types as adding or subtracting exactly 1, so the
conversion is exact for a constant 1 operand. Decimal already works
through the op_Increment/op_Decrement path.
Assisted-by: Claude:claude-fable-5:Claude Code
Optimized code stores no temporary for a deconstruction element that is
used only once after the deconstruction. MatchAssignments handled that
for trailing elements, but a nested deconstruction copies the inner
element to a temporary, so the elements preceding it are also left
without an assignment; their external load then violated the
DeconstructInstruction invariant that all pattern variable loads are
descendants of the instruction. The forwarding fixup now covers all
unassigned elements and inserts in pattern order, because the statement
and expression builders pair pattern variables with assignments
positionally. This also fixes the nested tuple deconstruction crash
reported in #3388.
Also unwrap the address of the tested operand in
VisitDeconstructInstruction: deconstructing a struct passes the
receiver by reference, which was emitted as an invalid cast,
'var (x, y) = (S)(ref s);', even without nesting.
Fixes#3388.
Assisted-by: Claude:claude-fable-5:Claude Code
Widening the compiler matrix answers the open review question on the
Issue3230 fixture guard empirically: a class naming its own nested
interface in its base list is a Roslyn-era relaxation. The legacy csc
rejects every such shape with CS0146 (circular base class dependency,
it never reaches the accessibility check), mcs 2.6.4 rejects them with
CS0122/CS0146, and mcs 5.23 accepts them all, rejects naming a base
class's protected nested interface with the same CS0122 as Roslyn, and
emits the same transitive InterfaceImpl metadata. The fixtures are
therefore gated to ROSLYN || MCS5, which exercises the base-list filter
on mcs-generated metadata as well. The pre-existing class C needs
MCS2-specific expected output because mcs 2.6.4 reorders interface-impl
rows and explicit implementations in metadata.
Assisted-by: Claude:claude-fable-5:Claude Code
A class may name its own protected nested interface in its base list
(class F : F.IFoo), but referencing a protected interface nested in a
base class there (class SubF : F, F.IFoo) does not compile, even though
that interface is accessible inside the class body. The interface-impl
metadata still lists such interfaces (they are inherited through other
entries), so emitting every entry produced uncompilable sources. Skip
base types that are neither nested within the current type's nesting
chain nor accessible from the enclosing scope without the
protected-through-inheritance privilege. The check covers every type
the base-list reference names: type arguments, array and tuple
elements, and the declaring chain of the named type.
Assisted-by: Claude:claude-fable-5:Claude Code
TypeDefTableTreeNode and MethodTableTreeNode carried their own copies of
the FieldList/MethodList/ParamList width rule, and ImplMapTableTreeNode
and FieldMarshalTableTreeNode derived heap index widths from the heap
size; both rules diverge from SRM (the *Ptr row count only decides
whether the pointer table itself is large, and heap widths are declared
by the HeapSizes flags, not derived). Rather than fixing the same bugs
in two places, these four nodes now consume the shared readers, which
also verify their computed widths against the SRM row size. The
remaining raw-reading nodes only use plain simple-index widths and are
left for a follow-up.
Assisted-by: Claude:claude-fable-5:Claude Code
The GUI has the metadata-tables view; the CLI had nothing, so checking
e.g. which MethodSemantics rows reference a Property row required a
hand-written System.Reflection.Metadata script. --dump-table <name>
prints every row of a table (RID, token, resolved names, heap offsets,
coded indexes) as an aligned text table, or as JSON with --json, for
the same 39 Cor tables the GUI shows.
Row enumeration for tables without public SRM row access lives in
MetadataExtensions next to the existing GetMethodSemantics helper, so
the GUI's raw-reading table nodes can be folded onto the shared
readers later. Every table's columns are spelled out explicitly in
ECMA-335 declaration order: reflecting over the SRM row structs would
tie the output (and its column order) to runtime internals, and
deterministic output is the point of the feature. JSON uses
System.Text.Json from the shared framework, so no new package
reference is needed.
Assisted-by: Claude:claude-fable-5:Claude Code
The Semantics and Association columns were read from offsets relative
to the metadata root instead of the current row: Semantics always
decoded the first two bytes of the metadata header and Association a
constant offset near it, so only the Method column ever carried real
row data. The Association coded-index width also used the plain-index
threshold (2^16) instead of the coded one (2^15 for one tag bit).
Rewrite the loop with a BlobReader positioned at the table start,
reading each column in sequence, and introduce SimpleIndexSize and
CodedIndexSize helpers encoding the ECMA-335 II.24.2.6 width rules.
The GUI's MethodSemantics metadata table view consumes this helper and
displayed the garbage values.
Assisted-by: Claude:claude-fable-5:Claude Code
The interactive view intentionally shows verbatim metadata names
(EscapeInvalidIdentifiers only runs for save/export), so names compiled
from F# or obfuscators can contain apostrophes. The Char span in the C#
highlighting grammar opened at any apostrophe, coloring everything up to
the next one as a char literal. A negative lookbehind limits the span to
positions where a real char literal can occur.
Assisted-by: Claude:claude-fable-5:Claude Code
The resolver comments cited section numbers from the C# 4.0 spec (and a
few from C# 9.0 drafts), which no longer match the published ECMA-334
standard. Renumber them against dotnet/csharpstandard draft-v11; every
reference was checked against the actual section headings. The old
'better conversion from type' subclause (7.5.3.4) no longer exists as
such and its rules live in 12.6.4.5-12.6.4.7, so that comment now says
so instead of pointing at a dead number.
Assisted-by: Claude:claude-fable-5:Claude Code
CI versions derived from branch names can contain underscores (e.g.
'avalonia12_1_1'), which dpkg rejects in the Version field. Map every
character outside dpkg's allowed set to '.' after the existing '-' to
'~' substitution.
Assisted-by: Claude:claude-fable-5:Claude Code
The tab-opening test only verifies document tab and selection wiring. Using a tiny in-assembly fixture avoids cold framework decompilation work in CI and lowers the chance of unrelated timeout noise.
Assisted-by: OpenCode:openai/gpt-5.5:OpenCode
Right-clicking inside the details editor inherited the metadata grid's
cell-oriented context menu, whose Copy entries are enabled only for a
hovered DataGridCell -- inside the details area there is none, so the
menu showed permanently disabled entries and selected text could not be
copied by mouse. The editor now carries the decompiler view's editor
menu shape: Copy (rich HTML copy with plain fallback, enabled while a
selection exists) and Select All. Enablement is decided in
ContextMenu.Opening, which only the context-request gesture raises;
the test therefore raises ContextRequested instead of calling Open().
Assisted-by: Claude:claude-fable-5:Claude Code
A one-line payload (short hex dump, tiny source-link document) rendered
as a squeezed strip barely taller than the row itself, which does not
read as an expandable details area. Floor the editor at 100px so the
details region stays visually recognisable regardless of payload size.
Assisted-by: Claude:claude-fable-5:Claude Code
The text-blob editor in the metadata row-details area hardcoded its font
and lacked the text view's flat selection highlight, so embedded source
looked different from the decompiled code right next to it and ignored
the user's font choice in the Options page. The decompiler-view look
(user-selected font applied live while attached, themed background,
square-cornered translucent selection) now lives in DecompilerTextEditor
itself, giving every surface hosting the editor the same appearance by
construction; DecompilerTextView and BuildTextBlob drop their now
redundant per-site styling.
Assisted-by: Claude:claude-fable-5:Claude Code
The CustomDebugInformation details area rendered decoded text payloads
(embedded source, source-link JSON, hex dumps) in a plain TextBox, which
shows code without any highlighting and materializes the whole formatted
text up front, so large embedded-source documents were expensive. Text
payloads now travel as a TextBlobDetail tagged with a file extension --
".json" for source link, the parent document's extension for embedded
source, none for hex -- and render in the theme-aware AvaloniaEdit
editor, which colours them via the existing highlighting registry and
virtualizes long documents. The editor's ThemeChanged subscription moves
from the constructor to OnAttachedToVisualTree so it stays paired with
the detach-time unsubscribe now that editors can leave and re-enter the
visual tree inside recycled row-details containers.
Assisted-by: Claude:claude-fable-5:Claude Code
The details content was pinned left and the text blob capped at 800px,
leaving dead space to the right of embedded-source text and the
flags/typed sub-grids. Let all three detail shapes stretch and give the
last sub-grid column the leftover width so the details area fills its
host row.
Assisted-by: Claude:claude-fable-5:Claude Code
The dotnet-hosted Roslyn 2.10 build cannot start its VBCSCompiler server
under a current dotnet host, so with /shared every test compilation first
waited out the client's full 20-second new-server connection timeout
before falling back to a sub-second in-process compile. Since the 2.10
configurations were enabled on non-Windows (#3914), that added ~29
minutes to the Linux CI job and ~43 minutes on macOS: ~340 affected
tests at ~21s each, versus ~0.2s for the toolsets whose server works.
Assisted-by: Claude:claude-fable-5:Claude Code
Review follow-up. A display-class field initialized from a non-this
parameter is now the only shape where propagation and a later mutation
coexist; it stays sound only because ResolveVariableToPropagate accepts
a parameter with LoadCount == 1, so the mutation can be redirected to
it. Nothing covered that, so Test12 pins it, and Test13 records the
neighbouring shape where the mutation happens inside a lambda - there
capturing the display class keeps it materialized and propagation never
arises. The guard predicate is renamed to say what it matches, since
'ReadOnly' reads like the C# keyword rather than 'a plain read'.
Assisted-by: Claude:claude-fable-5:Claude Code
A field that is propagated to the variable it was initialized from is
replaced by that variable, so re-emitting its initializer assigns the
variable to itself. Where the field was initialized from 'this' the
result does not even compile ('this = this'). The store is dropped
instead, which is what VisitStObj already does for initializer stores
that are not part of an object-initializer block; the insertion position
has to be tracked separately from the loop index, because skipping a
store would otherwise push the following ones past the end of the block.
Assisted-by: Claude:claude-fable-5:Claude Code
Aggressive scalar replacement propagated a display-class field to its
source variable even when the field is mutated after initialization,
aliasing two distinct source-level variables (Test9: thisField and
this). Propagation is now cancelled when the field sees a second store
or its address escapes, but only for propagation targets that cannot
absorb the store: 'this' and variables that are themselves
scalar-replaced display classes. Parameters continue to propagate,
because their remaining uses are already restricted by
ResolveVariableToPropagate and a captured parameter mutated inside a
lambda (DelegateConstruction's Bug951) must keep mapping to the
parameter. Checking CanPropagate first also keeps the guard away from
Mono state-machine fields, whose VariableToDeclare is pre-bound to a
state-machine variable that Propagate(null) would discard.
Re-enables Test9 and adds Test10 covering the escaping-address variant
(Interlocked.Exchange(ref displayClass.thisField, ...)).
Assisted-by: OpenCode:openai/gpt-5.5:OpenCode
Assisted-by: Claude:claude-fable-5:Claude Code
The .NET 10 BCL ships static [Extension] classes that contain ordinary
nested types (e.g. XDocumentExtensions.XDocumentNavigable). Decompiling
such a nested type's member in isolation resolved the enclosing
container's ExtensionInfo, and DecompileBody then dereferenced the
missing extension-member mapping. A container without any extension
blocks now reports no ExtensionInfo at all, and ResolveExtensionInfo
applies a container's info only to members that actually belong to one
of its extension blocks.
Assisted-by: Claude:claude-fable-5:Claude Code
Avalonia windows default to ShowInTaskbar=true, so every owned modal
dialog (Open from GAC/NuGet feed/running process, Manage Assembly
Lists, Export Project, Create List, Set Target Framework) got its own
taskbar button, which is not standard Windows behavior. The assertion
and crash dialogs intentionally keep their buttons so an interrupted
session stays easy to find.
Assisted-by: Claude:claude-fable-5:Claude Code
ProDataGrid derives its scroll extent from the current scroll offset, so an
offset that is briefly too large inflates the extent, which permits a larger
offset again. A trackpad reaches that state within a few hundred sub-row
events: instrumented in the running app, a ~1500px list reported 8735px and
kept growing, the thumb collapsed to its minimum, and the end of the list ran
away from the user. A second defect slid the rows sideways by up to 10px - the
star-sized column is measured against the width including the vertical scroll
bar, leaving the grid convinced it has a scroll bar's worth of content to
reach.
Disabling the horizontal scroll bar, giving each grid its own
DefaultRowHeightEstimator, pinning RowHeight, forcing the vertical scroll bar
visible, and disabling scrolling on the template's inner ScrollViewer each
removed a symptom at most; none can break a loop that runs through the scroll
offset, and 12.0.4 is the newest release. A ListBox's virtualizing panel keeps
the extent a function of the items alone. The price is laying out the columns
here - so the header row and the item template have to be kept in step - and
losing sortable, resizable headers.
Assisted-by: Claude:claude-opus-5[1m]:Claude Code
A collection that fails once the session has been granted - most plausibly
the target exiting before the stop command reaches it - left the task copying
the trace connection behind. That connection is torn down on the way out, the
copy faults, nobody awaits it, and the finalizer hands it to
TaskScheduler.UnobservedTaskException, which this app reports as a crash: a
second report of a failure the dialog's error bar had already explained
correctly, minutes later and detached from the gesture that caused it. The
CollectTracing2 fallback walks the same path, so one dying target produced two
of them.
The teardown order is the substance of the fix. The session connection has to
go first, because after the failure nothing else will ever end the read the
copy is parked on; the drain second, so that its own failure is observed
rather than abandoned; and the half-copied trace last, so nothing is still
writing into it when it is dropped.
The regression test pins the invariant rather than the symptom, because the
symptom is transport-specific: a Windows named pipe reports an aborted
overlapped read as cancellation, and a cancelled task is never unobserved, so
only the unix transport can produce the crash at all. What holds everywhere is
that no drain may still be running once the failure path is done with it.
Assisted-by: Claude:claude-opus-5:Claude Code
A refused unix socket raises SocketException, which derives from
Win32Exception rather than IOException, so it escaped the filter meant to
skip one unreachable process and failed the whole concurrent enumeration
instead: a machine where any .NET process exits between the port scan and
the connect showed an empty list. The classification is now a named
predicate covering both transports.
Two dialog defects shared a shape - state left behind by a query nobody is
waiting for any more. Rebuilding the bound collection on every filter
keystroke made the grid drop its selection and write that null back,
discarding the assemblies of a process the new filter still matched; and the
branch taken when nothing is selected cleared no loading flag, while the
query it superseded was no longer allowed to, so the progress bar animated
over an empty pane. Relatedly, the two-second command budget expired into a
process listed with null metadata, which made a slow machine look like a
runtime that answered with nothing; it is longer now, and expiry names the
process it gave up on, since the only thing that ever reaches the far end of
that budget is a runtime which will never answer.
The test gaps are closed the same way: a real dynamic assembly pins the
in-memory classification, the managed-only assertion is stated as the
property instead of a list of native names to exclude, and the .NET
Framework path gets live tests. Those showed that a desktop CLR process
mostly reports NGen native images rather than the IL assemblies behind them,
which is now recorded as the fidelity gap it is.
Assisted-by: Claude:claude-opus-5:Claude Code
With a modern .NET app the executable in the process list is a native
apphost that carries no IL, so the path a user can see is precisely the
one a decompiler cannot open - ILSpy itself is an example. Ask the
runtime instead: since .NET Core 3.0 every CoreCLR process serves a
diagnostics endpoint that names its managed entry assembly and, via an
EventPipe rundown, every assembly it has loaded, including ones behind a
single-file bundle or with no file at all. The endpoint answers the same
way on Windows, Linux and macOS and needs no privileges beyond same-user,
which also makes it the only workable route on macOS, where native
process introspection is gated by SIP.
The protocol and the nettrace container it returns are implemented here
rather than taken from Microsoft.Diagnostics.NETCore.Client, so the
feature costs no new package reference; the reader is scoped to loader
rundown events and steps over everything else by size. The rundown asks
for the loader keyword alone: the runtime's default set also collects the
JIT and IL-to-native-map rundown, which in a long-running process buries
the module list and overruns the session buffer, costing the very events
the dialog needs.
Windows additionally lists .NET Framework processes, which predate the
endpoint and are read from their OS module list instead.
Assisted-by: Claude:claude-fable-5:Claude Code
The decompiler suite is cross-platform code already exercised by the
Windows and Linux jobs; running it a third time on macOS only made the
slowest job of the matrix (~54 min vs ~41 min on Linux) slower without
adding coverage.
Assisted-by: Claude:claude-fable-5:Claude Code
The gate job was copy-pasted into build-ilspy.yml and build-frontends.yml
and the two copies had already started drifting (job casing, comment
wording). A workflow_call workflow keeps a single definition; the callers
shrink to a uses: job and route its "run" output to their build jobs
unchanged. Each caller must still grant pull-requests: read explicitly,
because the called workflow's token is the intersection of what the
caller grants and what the callee requests, and both callers default to
contents: read only.
Assisted-by: Claude:claude-fable-5:Claude Code
A branch with an open PR was built twice per push, once by the push
trigger and once by the pull_request trigger, doubling the Actions cost
of every PR iteration. The pull_request run is the canonical one (it
tests the merge with the base branch), so a small gate job now asks the
API whether the pushed branch has an open PR whose base would fire the
pull_request trigger and skips the push build if so. Pushes to master
and release/* are exempt because their runs publish packages; PRs
targeting other branches are ignored because they never trigger a
pull_request build that could replace the push one.
Filtering fork PRs in the on: section instead was considered and
rejected: trigger-level filters cannot see whether an open PR exists,
and skipping the pull_request runs would have lost merge testing.
Assisted-by: Claude:claude-fable-5:Claude Code
A bare `bash` on Windows can resolve to the WSL launcher
(System32\bash.exe) ahead of Git Bash, running the pre-commit hook
inside a Linux distro. There the CRLF-checked-out script fails to parse
("$'\r': command not found"), and even with clean line endings the WSL
path would install a Linux dotnet-format instead of the Windows one.
Resolve Git Bash from git.exe's install location instead, the same
CRLF-tolerant shell git itself uses to run the hook; non-Windows
platforms keep using plain bash.
Assisted-by: Claude:claude-fable-5:Claude Code
Each fixture test assembled Documentation/IdStringProbe.il separately,
and ilasm writes its output next to the input: on Windows the second
test failed with a sharing violation (0x80070020) because the first
test's PEFile still had the previous output open. The probe is now
assembled once per test run from a temp copy of the .il, so the tests
share one PEFile and nothing is written into the source tree, which
also makes the per-directory .gitignore unnecessary.
Assisted-by: Claude:claude-fable-5:Claude Code
IdStringMemberReference was its only implementation and left with the
ID string grammar parser; unlike the NRefactory-era type system, IMember
does not extend IMemberReference here, so nothing in the library
produces or consumes it anymore.
Assisted-by: Claude:claude-fable-5:Claude Code
A bound generic argument list without its closing brace fell off the
end of the string and silently produced an arity, making malformed ID
strings appear to parse and resolve to nothing instead of failing with
the documented ReflectionNameParseException. Also replaces the non-ASCII
punctuation in comments added by this branch with ASCII equivalents,
per the repository convention. Both raised by review on #3926.
Assisted-by: Claude:claude-fable-5:Claude Code
The C#/Roslyn-form ID of a member whose C++/CLI form differs can equal
the only key of a same-named sibling overload (char* vs signed char*).
Assemblies containing such overloads cannot come from the C# compiler,
so their xml files use the C++/CLI dialect, where that key documents
the sibling: falling back to the Roslyn form would show the sibling's
documentation for an undocumented member. GetIdStringCandidates now
omits the Roslyn form when a same-named sibling's C++/CLI form owns it,
so a lookup miss stays a miss.
Assisted-by: Claude:claude-fable-5:Claude Code
The recursive-descent parser for the full ID string grammar existed only
to feed the type-system-based FindEntity, while signatures were already
matched by regenerating candidate IDs; with two dialects a parser would
have to implement both grammars and stay in sync with the generator.
FindEntity now only decodes the structural skeleton - the declaring type
name and the member name - resolves the type, and narrows the members by
metadata name before the generate-and-compare step, falling back to an
unfiltered scan for keys whose name does not equal the metadata name
(a C++/CLI 'default' indexer). ParseTypeName, ParseMemberIdString and
their reference types (IdStringMemberReference,
GetPotentiallyNestedClassTypeReference) are removed.
Assisted-by: Claude:claude-fable-5:Claude Code
Member lookup compares regenerated ID strings, and with two dialects a
single mixed pass can resolve to the wrong member: the stripped
C#/Roslyn form of one member can equal the C++/CLI-dialect key of a
different member, e.g. C++ overloads differing only in a custom
modifier such as char* vs signed char*. FindMemberInType therefore
runs two passes over the whole member list, the more specific C++/CLI
form first, so an MSVC-written cref resolves to the member it names
instead of the first member whose stripped form happens to collide.
Assisted-by: Claude:claude-fable-5:Claude Code
The member keys in an xml doc file depend on the compiler that wrote it
(C# vs the MSVC C++/CLI dialect), so XmlDocumentationProvider's
entity-based lookup now tries each candidate form in order. The tooltip
path goes through the entity overload instead of building the key
itself, so the fallback lives in one place. The C++/CLI form is tried
first: wherever it differs it contains character sequences Roslyn never
writes, so it can only match MSVC-generated keys, while the stripped
Roslyn form of one member could match the key of a different member in
an MSVC-generated file.
Assisted-by: Claude:claude-fable-5:Claude Code
MSVC documents C++/CLI members with ECMA-372-style ID strings that
differ from Roslyn's in signatures: custom modifiers are rendered after
the modified type (a 'const int' parameter becomes
System.Int32!System.Runtime.CompilerServices.IsConst), arity markers
stay on generic instantiations (List`1{System.Int32}.Enumerator), and
the default indexed property is called 'default'. Roslyn ignores
modifiers entirely, so one generated string cannot match both
compilers' xml files: GetIdString keeps producing the C#/Roslyn form,
and the new GetIdStringCandidates additionally yields the C++/CLI form,
most specific first, for lookup code to try in order.
The dialect is pinned by IdStringProbe.il/.xml, the trimmed disassembly
of an MSVC-compiled probe assembly together with the unmodified xml MSVC
generated for it. Notable observed deviations from MSVC's documented
format: modreq is generated, but only modreq(IsVolatile) uses the
documented '|'; modreq(IsByValue) on conversion operator operands is
rendered with '!'.
Assisted-by: Claude:claude-fable-5:Claude Code
FindEntityInRelevantAssemblies now uses the metadata-level FindEntity
instead of parsing the ID string into type-system references and
resolving them per assembly. Two behaviors of the old path need
explicit handling because FindEntity only searches the modules it is
handed: reference assemblies are skipped so the search prefers an
assembly with a usable definition, and a member whose declaring type is
present only as a type forwarder is looked up in the assembly the
forwarder points to, which the assembly resolver may load on demand
(the old path got this through DecompilerTypeSystem resolution).
Assisted-by: Claude:claude-fable-5:Claude Code
Resolve an ID string to a (module, handle) pair by scanning metadata
directly: namespace/type-name splits are tried at every dot (the format
does not mark the boundary), nested types and type forwarders are
walked, and members are matched by regenerating each candidate's ID
string instead of parsing the signature portion, which keeps resolution
in sync with generation by construction. This gives navigation a
resolution path that needs no type system and works on assemblies
exactly as their metadata spells them. Inherent format limitations
(metadata names containing ID string special characters; function
pointer types rendering empty, so such overloads share an ID) are
documented on the method.
Assisted-by: Claude:claude-fable-5:Claude Code
Generate ID strings directly from metadata instead of the type system,
so callers holding only a MetadataFile and an EntityHandle do not need
to materialize a compilation first, and raw metadata names survive
verbatim. GetIdString(IEntity) stays as a thin wrapper over the new
implementation, keeping the published entity-based entry point intact.
The implementation is validated by a differential suite comparing every
symbol of a test corpus against Roslyn's GetDocumentationCommentId.
Corners pinned by those tests: generic arguments distribute to their
nesting level (Outer{A}.Inner{B}), op_CheckedExplicit carries the
~ReturnType suffix like the other conversion operators, and custom
modifiers are ignored like Roslyn ignores them (a virtual method's 'in'
parameter is modreq(InAttribute) but documented as T@). Array shapes
use the spec's lowerbound:size notation, covered by a hand-built
module, since C# cannot express non-default array bounds in signatures.
Assisted-by: Claude:claude-fable-5:Claude Code
dependabot[bot]
Siegfried Pammer
Christoph Wille
ds5678