This is a cherry pick of the 3.1 PR.
https://dev.azure.com/dnceng/internal/_git/dotnet-aspnetcore/pullrequest/9042
The only change was to remove a questionable call to Abort after catching the exception.
There are some less severe issues in Kestrel and MVC that we'll follow up on publicly after this releases.
* Remove all ref/ projects
* Remove GenAPI infrastructure
* Remove notion of a reference assembly project
- remove `$(IsReferenceAssemblyProject)`, `$(ReferenceReferenceAssemblies)` and `$(ReferenceImplementationAssemblies)`
- remove unnecessary `$(NoWarn)` settings
nits:
- remove a few misleading comments
- wrap some long lines
* Move .0 package version workaround into Versions.props
- touch up SharedFramework.External.props
* Expose `%(LatestPackageReference.RTMVersion)` metadata
- automate use of properties in the `@(LatestPackageReference)` item group to make this maintainable
- add a couple of special cases at the bottom of eng/Dependencies.props
- add one more `$(...PackageVersion)` property to avoid yet-another special case
* Enable Roslyn reference assemblies
- exclude ref/ assembly from packages other than targeting pack
- update Microsoft.AspNetCore.App.Ref.csproj
- `%(IsReferenceAssembly)` and `%(ReferenceGrouping)` metadata no longer relevant
- only ref/ assemblies are in `@(ReferencePathWithRefAssemblies)` item group
nits:
- remove now-unnecessary workaround
- issues with TFM transition are behind us
- clean up Microsoft.AspNetCore.App.Runtime.csproj slightly
- use `GeneratePathProperty="true"`
- reorder item / property settings for meta-expansion
- correct spelling errors and phrasing in comments
* Update documentation to reflect recent changes
- remove CrossRepoBreakingChanges.md; was tied to old TeamCity infrastructure
- also much less relevant given repo merges
- adjust details and examples in ReferenceResolution.md
- reflect repo merges, Dependencies.props changes, and current Maestro++ channels
- add a few more details e.g. specific files where Version.Details.xml versions are used
* !fixup! Remove another irrelevant doc file
* !fixup! Address PR review suggestions
- convert a couple of warnings to errors
- use consistent casing for Microsoft.NETCore.App.Runtime.* packages
- reduce `%(LatestPackageReference.Version)` metadata special cases
- add and improve comments e.g.
- improve comments about `$(*V0PackageVersion)` properties
- improve placement of comments about item removal in ResolveReferences.targets
- confirmed `$(*V0PackageVersion)` property list is complete
nits:
- fix solution example in ReferenceResolution.md
- remove item group definition for `@(LatestPackageReference)`
- remove `%(LatestPackageReference.VersionName)` metadata after use; large item group
- similarly, remove `%(LatestPackageReference.RTMVersion)` when not needed; just complicates `Condition`s
When I squash, I must remember this fixes
- #14801
- dotnet/aspnetcore-internal#2693
* Actually use `%(LatestPackageReference.RTMVersion)` metadata
- gather RTM package references in a new project
- a (very) separate project to work around package conflict resolution
- empty `Test` target works around Arcade's testing approach
- new target in ResolveReferences.targets updates relevant assembly paths to use the RTM packages
- done as soon as possible after `ResolvePackageAssets` determines the paths
- done for all compilation inputs, not just ref/ assemblies
All rules are currently disabled, except for one that I enabled for
src/Http via a new ".editorconfig" file I added there.
Other changes:
* Allow editorconfigs in MVC and Razor to flow to the root
* Consolidate a few editorconfig settings
* Tweak Ruleset config in Azure/AzureAD where it clashed.
Addresses the beginning of #9620, but it's a fair chunk of work to
enable most rules through the whole repo. That can be done directory by
directory and rule by rule by dropping .editorconfig files though.
* Use preferred overloads of string.Split
* Revert TrimEntries for netstandard2.0
Revert usage of StringSplit.TrimEntries for projects that target netstandard2.0.
Co-authored-by: Levi Broderick <levib@microsoft.com>
* nit: Remove useless `$(HasReferenceAssembly)` settings
- set in /Directory.Build.targets
- `true` only in `$(IsAspNetCoreApp)` projects
* nit: Remove useless `$(CompileUsingReferenceAssemblies)` settings
- no current versioning differences between ref/ and src/ assemblies when targeting default TFM
* Add more `$(GenerateDocumentationFile)` settings
- increases the number of generated doc files, mostly without problems
- !fixup! correct typo in `DebugProxyHost` doc comments
- was not generating a doc file before
- remove previous (ineffective) src/Components/Directory.Build.targets setting
- nit: remove a duplicate `$(GenerateDocumentationFile)` setting
* nit: Remove useless `$(IsPackable)` settings
- only analyzers and implementation projects are packable by default
- main use case for explicit setting is projects shipping only in shared framework
- conditional setting in src/Mvc/Directory.Build.props just subset logic in /Directory.Build.targets
* nit: Remove useless `$(IsProjectReferenceProvider)` settings
- only implementation projects are providers by default
* nit: Remove useless `$(IsTestAssetProject)` settings
- set in src/Mvc/test/WebSites/Directory.Build.props
* !fixup! Looks like `InProcessNewShimWebSite` must compile w/o ref/ assemblies
- restore `$(CompileUsingReferenceAssemblies)` in this one project
* Remove extra `[SkipOnHelix]` attribute
- only need the one compiled into Microsoft.AspNetCore.Testing
- update the documentation to reflect this
- nit: address Markdown warnings that VS Code showed
* Add `SuccessfulTests` to ensure something runs in every non-Helix runs
- #22241
- cleans up hundreds of warnings but leaves a couple for the Blazor tests assembly
- see comments about xUnit runner command line in the new class
* Improve build reliability
- ensure `ResolveCustomReferences` target executes before packages are used
- `ResolveAssemblyReferences` and `ResolveAssemblyReferencesDesignTime` targets run too late
- e.g. failed builds of Microsoft.AspNetCore.WebUtilities or Microsoft.AspNetCore.Hosting when building from root
- add `GetReferenceProjectTargetPathMetadata` for ease of use as well as reliability
- avoids extra work to get existing metadata (ref/ projects execute no tasks in this target)
nit: rename `@(ReferenceProjectMetadata)` -> `@(ReferenceProjectTargetPathMetadata)`
* Ensure `GetTargetPathMetadata` target runs with `$(TargetFramework)` set
- ref/ projects all multi-target and otherwise no-op this target
* Revert "Fix various "Type or namespace not found" errors (#20736)"
- change is no longer needed with other fixes in this PR
This reverts commit 8218d6e0e7.
* Fix use of precedence in endpoint routing DFA
Fixes: #18677Fixes: #16579
This is a change to how sorting is use when building endpoint routing's graph of
nodes that is eventually transformed into the route table. There were
bugs in how this was done that made it incompatible in some niche
scenarios both with previous implementations and how we describe the
features in the abstract.
There are a wide array of cases that might have been impacted by this
bug because routing is a pattern language. Generally the bugs will involve a
catch-all, and some something that changes ordering of templates.
Issue #18677 has the simplest repro for this, the following templates
would not behave as expected:
```
a/{*b}
{a}/{b}
```
One would expect any URL Path starting with `/a` to match the first
route, but that's not what happens.
---
The change supports an opt-in via the following AppContext switch:
```
Microsoft.AspNetCore.Routing.UseCorrectCatchAllBehavior
```
Set to true to enable the correct behavior.
---
The root cause of this bug was an issue in how the algorithm used to be
build the DFA was designed. Specifically that it uses a BFS to build the
graph, and it uses an up-front one-time sort of endpoints in order to
drive that BFS.
The building of the graph has the expectation that at each level, we
will process **all** literal segments (`/a`) and then **all** parameter
segments (`/{a}`) and then **all** catch-all segments (`/{*a}`). Routing
defines a concept called *precedence* that defines the *conceptual*
order in while segments types are ordered.
So there are two problems:
- We sort based on criteria other than precedence (#16579)
- We can't rely on a one-time sort, it needs to be done at each level
(#18677)
---
The fix is to repeat the sort operation at each level and use precedence
as the only key for sorting (as dictated by the graph building algo).
We do a sort of the matches of each node *after* building the
precedence-based part of the DFA, based on the full sorting criteria, to
maintain compatibility.
* Add test
* Made changes to FileBufferWriteStream
- Make the internal FileStream write only
- Make a new readable stream over the same file in DrainBufferAsync to copy data to the buffer.
- Added an overload to DrainBufferAsync into a PipeWriter and use this overload in the 2 formatters in MVC. This should reduce the amount of copying from the internal buffer and reduces pinning (since these buffers are already pinned)
- Improved formatter tests
* Fix use of precedence in endpoint routing DFA
Fixes: #18677Fixes: #16579
This is a change to how sorting is use when building endpoint routing's graph of
nodes that is eventually transformed into the route table. There were
bugs in how this was done that made it incompatible in some niche
scenarios both with previous implementations and how we describe the
features in the abstract.
There are a wide array of cases that might have been impacted by this
bug because routing is a pattern language. Generally the bugs will involve a
catch-all, and some something that changes ordering of templates.
Issue #18677 has the simplest repro for this, the following templates
would not behave as expected:
```
a/{*b}
{a}/{b}
```
One would expect any URL Path starting with `/a` to match the first
route, but that's not what happens.
---
The root cause of this bug was an issue in how the algorithm used to be
build the DFA was designed. Specifically that it uses a BFS to build the
graph, and it uses an up-front one-time sort of endpoints in order to
drive that BFS.
The building of the graph has the expectation that at each level, we
will process **all** literal segments (`/a`) and then **all** parameter
segments (`/{a}`) and then **all** catch-all segments (`/{*a}`). Routing
defines a concept called *precedence* that defines the *conceptual*
order in while segments types are ordered.
So there are two problems:
- We sort based on criteria other than precedence (#16579)
- We can't rely on a one-time sort, it needs to be done at each level
(#18677)
---
The fix is to repeat the sort operation at each level and use precedence
as the only key for sorting (as dictated by the graph building algo).
We do a sort of the matches of each node *after* building the
precedence-based part of the DFA, based on the full sorting criteria, to
maintain compatibility.
* Add test
* Fixing QueryCollection property documentation
* Update src/Http/Http/src/QueryCollection.cs
Co-Authored-By: Chris Ross <Tratcher@Outlook.com>
Co-authored-by: Sourabh Shirhatti <shirhatti@outlook.com>
Co-authored-by: Chris Ross <Tratcher@Outlook.com>
* Don't re-use DefaultHttpContext if IHttpContextAccessor is in use
- Consumers may still get null or an ODE but will never end up with data from a different request.
- Make sure an ODE is thrown from all properties on HttpContext after the request is over.
**Description**
An infinite loop can happen in routing if there is a catch all route with host name matching.
This problem is caused by the DFA matcher builder giving an incorrect exit destination to policies. Currently the exit destination is the catch all state, so the policy will transition to itself when there is no match. It will run again, transition to itself again, run again, etc. This causes the policy to run forever.
What should happen is the host name policy fails, it transitions to the final state with no candidates, and the route matcher does not match any endpoints. The browser is returned a 404 status.
**Customer Impact**
This problem shows up in this situation:
1. If a customer has configured a catch all route in their app
2. The catch all route has host matching
3. A browser makes a request to the server that matches the catch all route but doesn't match the host name
The route matcher will run forever, using up a threadpool thread. When threadpool threads are exhausted the server will stop responding.
**Regression?**
No.
**Risk**
Medium. The fix is simple but route matching is complex, and routing runs with every request.
Chris Ross (ASP.NET)
Doug Bunting
Kevin Pilch
Pranav K
Martin Costello
Stephen Halter
Chris Ross
James Newton-King
dotnet-maestro[bot]
Filip Staffa
ZeBobo5
Ondřej Štorc
Hao Kung
msftbot[bot]
Kahbazi
Justin Kotalik
John Luo
N. Taylor Mullen
Ryan Nowak
David Fowler
Sébastien Ros
dotnet-bot
William Godbe
Thomas Levesque
Alessio Franceschelli
Chris Oliver
James Newton-King
Kevin Pilch