Data-annotations length fix #24101

weitzhandler · 2017-09-17T00:05:04Z

Fixes #24064
See this comment.

…ection<T>

- Linux x64 Release Build - UWP CoreCLR x64 Debug Build

- Extracted Count checking to an external helper to obey DRY - Removed dependency of ICollection<T> and changed to simple reflection Count property lookup

weitzhandler · 2017-09-17T04:04:44Z

Looks like all tests passed.

danmoseley · 2017-09-18T17:38:18Z

...m.ComponentModel.Annotations/src/System/ComponentModel/DataAnnotations/MaxLengthAttribute.cs

+            {
+                property = value.GetType().GetRuntimeProperty("Count");
+            }
+            catch (TypeAccessException)


As there aren't other uses of this pattern in this library, could you please add a comment like, "// On CoreRT, this property may not be enabled for reflection."

Done 6170e27.

danmoseley · 2017-09-18T23:16:42Z

Thanks @weitzhandler !

weitzhandler · 2017-09-18T23:21:34Z

@danmosemsft yet I forgot to add the enclosing parenthesis to my comment...
Not that I think anyone cares 😝

ajcvickers · 2017-09-18T23:22:45Z

This fix seems wrong to me because it changes the behavior when .NET Native is being used such that a nonsensical exception will be thrown which would never be thrown when .NET Native is not being used. These kinds of errors are really hard for developers to debug.

@atsushikan @danmosemsft Given the limitations in .NET Native, should we avoid adding code that uses Reflection in new .NET Core code going forward? (Btw, this is truly a case where the type is completely unknown to the framework can could be almost anything.)

If we are going to continue to use Reflection in new .NET Core code, then should it be the case that:

The runtime gives enough details to the application so the developer can correctly figure out what to add to rd.xml to get things working?
- If the app developer doesn't do this, then it will continue to not work
Code that uses Reflection should provide helpful message as to what is not supported on .NET Native?
- That is, it won't just fail in some non-obvious way like in this PR.

weitzhandler · 2017-09-18T23:26:10Z

@ajcvickers
We could for now rethrow the exception, maybe wrapped as an inner exception to a more familiar dedicated exception adding some details.

danmoseley · 2017-09-19T00:33:15Z

@atsushikan should probably answer that. I haven't seen a case before where we were reflecting over a type we didn't control, but surely cases must exist.

danmoseley · 2017-09-19T00:33:26Z

@weitzhandler yeah I wasn't going to reset CI just for a paren :)

danmoseley · 2017-09-19T00:34:15Z

@ajcvickers apologies for merging code without you - I saw that Atsushi had signed off, and read that as your name instead.

ghost · 2017-09-19T14:30:52Z

should we avoid adding code that uses Reflection in new .NET Core code going forward?

You should avoid unnecessary use of Reflection when an easy alternative is available (a good example being using Type.Name in exception messages instead of Type.ToString()), or using Reflection to get the members of enums which are internal to your library and thus, under your control.

Some libraries are inherently dynamic in nature and won't be a easy fit for .NET Native ever. We still let the determined developer try to use them but realistically, he'll be spending lots of time getting these exceptions from the runtime and modifying their rd.xml and ending up with a lot of the very bloat that .NET Native is supposed to eschew...

danmoseley · 2017-09-19T16:21:06Z

@ajcvickers would you like me to simply revert this, or is there a modification you would like?

ajcvickers · 2017-09-19T16:45:48Z

@danmosemsft I would like to discuss with @divega. I suspect given the answer from @atsushikan that we will need to revert the entire improvement since I can't think of any way to do it without Reflection, and using Reflection will be a bad experience on .NET Native. But maybe we can come up with something reasonable.

Btw, I'm really surprised by the idea that you guys don't see Reflection on arbitrary types used more often. For example, EF uses it extensively, and EF is installed in a large percentage of Visual Studio projects.

danmoseley · 2017-09-19T16:51:12Z

you guys don't see Reflection on arbitrary types used more often.

I guess it was more that I hadn't seen mention of this problem before. Maybe .NET Native isn't high on their radar.

weitzhandler · 2017-09-19T22:01:48Z

Hi guys.
I have another idea.
As this all tunnels down to the ICountable issue #23578 which doesn't seem to be addressed so quickly, and HashSet<T> is the prominent type that is not countable in the .NET, why don't we just make HashSet<T> implement ICollection?
I opened a new issue for this: #24156.

I wish we are able to introduce the ICountable/IReadOnlyCollection (non-generic) interface anyway, because its lack is a design flaw. I'd be curious to see what the impact of implementing it would be, at least inside this repo.

ajcvickers · 2017-09-20T16:28:17Z

@weitzhandler Feel free to per peruse other ideas like you suggest, but that will take time even if they are eventually agreed upon. We can revisit this issue after those ideas pan out, but for now we need to get something in that works or back out what we have already done.

* Microsoft.ServiceModel.Syndication skeleton project * Adding the existing classes of SyndicationFeed from .net fx * Added the needed references to get the code to compile * Changed some namespaces * Fixed errors when reading feeds and replaced some buffers * Cleaning code for PR * Deleted some unused files * Added the posibility that the user creates his own date parser when reading a feed from outside, also as part of our default date parser, if it can't parse the date it will just assign a default date to avoid the crash * Added correct testnames and Copyright * Initial changes to add custom parsers * Added more delegates as parsers * Added custom parser delegates * Save changes * Initial SyndicationFeed * Removed the dependence of the old SR class * Cleaned the code, deleted Diagnostics, fixed throwing resources * Cleaned code from most of the unnecesary comments * Formatted the code with CodeFormatter Tool * Moved the call of itemParser to be called with each item * Test using custom parsers * Fixed image issue where image tag was using original title and url, for RSS formatting Image issue fixed and added some tests * Test clase jorge * Save changes with Jorge cass * Initial Jorge * saving changes * Save changes * Fixed image and items issue * Fixed disjoint items in atom * Run codeFormatter * Adding parsing to optional elements int the spec added unittesting * Added Icon parsing to Atom10 feedFormatter. Added unit test * Adding Async Writing RssFormater writes new optional spec elements * Added Icon writing on Arom Writer * Fixed some warnings * Improved custom parsing for RSS * Added custom parsing for Atom feed formatter, added test * added nameof() to all exceptions when needed. * Adding Extension Methods for XmlReader * Fixing code for PR * Fixing code for PR * Added check for skip days to allow only accepted days. * Improved flexibility for Default dateparser * Add wrong dates example * Fixing code review * Fixed warnings on some unawaited methods. * Added async extension methods for XmlReader * Add XmlWriterWrapper method extensions * Changed ReadCategoryFromAsync to return a SyndicationCategory * Fixed sync method exposed GetReader. * Edited XmlReaderWrapper, moved methods to extension methods. * Removed unnecesary variables from Wrappers. * Fixed bug from ServiceModel.Syndication * Make BigInteger ctor opt path for == int.MinValue reachable. The BigInteger(ReadOnlySpan<byte> value) ctor contains an optimised path for value being larger than 4 bytes and the result being equal to int.MinValue. However this path is inside a test that precludes that value, so can never be reached. Restructure so that the path is reachable. * Fix ServiceModel.SyndicationFeed project dependencies - Change M.SM.SyndicationFeed to be a .NET Standard 2.0 library - Change tests to use official .NET Core 2.0 release from preview * Add btrfs and other missing file system types (dotnet#24102) * Uppercase * Add more filesystems from stat.c. Change to stat names where possible * Add some more local friendly names * Add some more remote file types from stat.c * Add entry to switch present in enum * Build errors * Remove Fixed entries that should be RAM * comment * Change case of 0X to 0x * Move cramfs to Fixed * ConcurrentQueue 128-byte cache line (dotnet#22724) ConcurrentQueue 128-byte cache line * Add warning by default in SGEN (dotnet#24054) * Output warning by default if run the tool directly without /quiet parameter. * add quiet parameter in the command. * fix parameter error. * Update the warning. * Add the target to copy the serializer to publish folder. (dotnet#24096) * Wrap cert callback leaked exceptions in WinHttpHandler (dotnet#24107) Similar to the fix done for CurlHandler (dotnet#21938), fix WinHttpHandler so that leaked exceptions from the user-provided certificate callback are properly wrapped. The ManagedHandler still needs to be fixed since it is not wrapping properly. Contributes to #21904 * Ensure ProcessModule for main executable is first in modules list (dotnet#24106) * Disable NegotiateStreamTest fixture for distros without working Kerberos (dotnet#24098) Disable NegotiateStreamTest fixture entirely because setup-kdc.sh is broken on some distros * Expose and add tests for Guid.ctor(Span)/TryWriteBytes/TryFormat * Address remaining PR feedback * #24112 Replaced documentation summary of TryPop with text similar to TryDequeue. (dotnet#24113) * Wrap exceptions from ManagedHandler's server validation callback (dotnet#24111) To match other handlers' behaviors. * Fix libgdiplus function loading on OSX. * Prevent ServiceControllerTests from dispose when already disposed (dotnet#24042) * Disable ServiceProcessTest that has been failing on CI and official builds * Prevent dispose from been called when already disposed * Add logging to repro if RemoveService is been called twice on the same ServiceController * Data-annotations length fix (dotnet#24101) * Updated in MinLengthAttribute and MaxLengthAttribute to support ICollection<T> * Added tests * Fixed typo * Trying to address two failing checks: - Linux x64 Release Build - UWP CoreCLR x64 Debug Build * Implemented changes requested in review - Extracted Count checking to an external helper to obey DRY - Removed dependency of ICollection<T> and changed to simple reflection Count property lookup * Added requested tests * Added catch for MissingMetadataException. * Extracted code from try-catch. * Added comment as requested. * Typo correction * Remove System.Drawing.Common from the netfx compat package (temporary). * Updating corefx to use new roslyn compiler (dotnet#24076) * Updating corefx to use new roslyn compiler * Updating to new version of the compiler and use the switch so tests pass on desktop * Fix System.Reflection.Metadata tests * Stop running Math.Clamp tests on UAP as API is not there (dotnet#24125) * Stop running Math.Clamp tests on UAP as API is not there * Disable only for AOT * Fix instantiating PrincipleContext with Domain type (dotnet#24122) * Fix instantiating PrincipleContext with Domain type * Enhancement * Disable Test_Write_Metric_eventListener on UWP. (dotnet#24127) * Revert "Remove System.Drawing.Common from the netfx compat package (temporary)." This reverts commit f6b0fbd. * Update BuildTools, CoreClr, CoreFx, CoreSetup, Standard to prerelease-02019-01, preview1-25718-02, preview1-25718-03, preview1-25717-02, preview1-25718-01, respectively (dotnet#24075) * Fixed compile warning/error on FreeBSD (dotnet#24141) * Marking {ReadOnly}Span as readonly structs (dotnet#23908) * Marking {ReadOnly}Span as readonly structs, fixing issue #23809 * Adding readonly attributes to reference assemblies. * Using "readonly ref" keyword instead of attributes. * Adding a LangVersion 7.2 property * System.Drawing: Throw ArgumentNullException on Unix as well (dotnet#24140) * Throw ArgumentNullException when stream is null * Throw ArgumentNullException when stream is null * Remove invalid NameResolution tests (dotnet#24147) The Dns_GetHostEntryAsync_* are fundamentially invalid because it's possible for the broadcast address, 255.255.255.255, to have an DNS mapping via manually modifying the hosts file. This was actually happening on Mac systems as well as an virtual environment running on top of that (i.e. Windows on Parallels). Ref: https://github.com/dotnet/corefx/issues/23992#issuecomment-330250642 Contributes to #23992 * Bump system.drawing.common.testdata to 1.0.6 * Update BuildTools to prerelease-02019-02 (dotnet#24146) * Fix path to test data * Fix whitespace * Add GraphicsTests based on Mono's test suite. * Consolidate more code in the "System.Drawing" namespace. * Remove all remaining Win32 codepaths from the mono codebase. All of this code now implicitly assumes that it will be run on a Unix platform. * Consolidate the rest of the gdipFunctions.cs file into Gdip.cs and GdipNative.Unix.cs * Consolidate the GraphicsUnit and ImageType enumerations -- they were duplicated. * Remove the mono Status enum and use the Windows constants instead in all Unix code. * Move all files into the regular directory structure. Suffix them with ".Unix" and ".Windows" when there are collisions. * Tiny bit of code cleanup * Add conditionals for recent versions of mono * Remove duplicate tests. * Fix multiplying TextureBrush with a disposed matrix on Unix (dotnet#24109) * Fix multiplying TextureBrush with a disposed matrix on Unix * Enable another passing test * Fix accidentally committed file * Add an error code fixup to Bitmap.Unix.cs to match Windows. * Bump System.Drawing.Common.TestData to 1.0.6 (dotnet#24149) * Bump system.drawing.common.testdata to 1.0.6 * Fix path to test data * Fix whitespace * Cleanup - Add/simplify using statements of disposable resources (Graphics, Bitmap) * Validate HatchStyle passed to HatchBrush ctor * Delete accidentally duplicated HatchBrush tests in LinearGradientBrushTests * Remove references to historical Mono bug IDs * Renable some already passing tests * Remove Thread.Sleep workaround * Update BuildTools to prerelease-02020-01 (dotnet#24172) * Add thread-local based switch to opt-in to ManagedHandler * Address PR feedback * PR feedback * Fix memory map imports (dotnet#24176) * Fix memory map imports Imports lost the last error attribute. Add it back and change the results to be the more correct "bool". Tweak the usage based on the new return type. #24159 * Move spin wait * Remove unused FEATURE_RANDOMIZED_STRING_HASHING (dotnet#24178) * Adding System.Data.Odbc package and including in metapackage * Fix MultiplyTransform with a disposed brush * Fix for 2nd Connection hangs in Mirroring Environment (dotnet#24181) * Switch tests to use thread-local switch for ManagedHandler And as a result re-enable parallelism of the test suite, which on my machine reduces the running time of the outerloop tests from 150s to 45s. * Update CoreClr, CoreSetup, ProjectNTfs, ProjectNTfsTestILC, Standard to preview1-25720-03, preview1-25719-04, beta-25721-01, beta-25721-01, preview1-25721-01, respectively * CoreFx #22406 Span based APIs - Text Reader Writer (dotnet#23786) * [Drawing] Move remaining "Unix" files into regular directory structure * As part of this, the "Windows" version of ImageFormat.cs is now being used everywhere. This file just contains a bunch of GUID's that are not platform-specific in any way. * Change AsSpan to property Span and rename AsMemory to Memory * Add Metafile and MetaHeader tests based on Mono's System.Drawing unit tests * Update project file * MetafileTests: Remove duplicates, and re-enable tests on Unix which are now working. Delete duplicate MetaHeadertests * Rationalize using statements * Update project file * Fix Metafile exception behavior * Simplify tests, remove duplicate test * Don't catch an ArgumentException just to get a NullReferenceException instead. * Assert.False(Object.ReferenceEquals => Assert.NotSame * PR feedback * Remove duplicate tests, fix typo * Remove X11 dependency for tests which are enabled on Unix. * Workaround libgdiplus glitches Don't test metafileHeader.Bounds when using libgdiplus. Don't test metafile.GetBounds on Unix, force MetafileHeader.Bounds to return an empty rectangle when MetafileSize == 0 Don't validate graphicsUnit on Unix. * Move Syndication to root/src * Disable Build of S.SM.Syndication. * Increase file descriptor limit in S.N.Http tests on macOS * Expose/test String.Create span-based method (dotnet#23872) * Expose/test String.Create span-based method * Address PR feedback * Update build to clang/llvm 3.9 (dotnet#24177) Update scripts, docs and build pipeline docker images to clang/llvm/lldb 3.9 * Update ProjectNTfs, ProjectNTfsTestILC, Standard to beta-25722-00, beta-25722-00, preview1-25722-01, respectively (dotnet#24207) * Remove the line that will copy the generated serializer to the pack. (dotnet#24199) * Revert "Update build to clang/llvm 3.9 (dotnet#24177)" This reverts commit 21e008a. * Remove stale SetStateMachine call in test * Ssl Stream Async Write (dotnet#23715) * Change from APM to Async/Await for write side * Added nameof * Reacting to review * Reacting to review * SSLStream : Fixed spelling mistake in file name (extra a) (dotnet#24221) * Fixed spelling mistake in file name * Fix csproj * Update ProjectNTfs, ProjectNTfsTestILC to beta-25723-00, beta-25723-00, respectively (dotnet#24222)

* Updated in MinLengthAttribute and MaxLengthAttribute to support ICollection<T> * Added tests * Fixed typo * Trying to address two failing checks: - Linux x64 Release Build - UWP CoreCLR x64 Debug Build * Implemented changes requested in review - Extracted Count checking to an external helper to obey DRY - Removed dependency of ICollection<T> and changed to simple reflection Count property lookup * Added requested tests * Added catch for MissingMetadataException. * Extracted code from try-catch. * Added comment as requested. * Typo correction Commit migrated from dotnet/corefx@b4daeab

Shimmy Weitzhandler added 7 commits August 30, 2017 02:05

Updated in MinLengthAttribute and MaxLengthAttribute to support IColl…

8650ca0

…ection<T>

Added tests

7d23869

Fixed typo

86ace1d

Trying to address two failing checks:

dfb5e1d

- Linux x64 Release Build - UWP CoreCLR x64 Debug Build

Implemented changes requested in review

6353a8f

- Extracted Count checking to an external helper to obey DRY - Removed dependency of ICollection<T> and changed to simple reflection Count property lookup

Added requested tests

4cba646

Added catch for MissingMetadataException.

409ee6f

dnfclas added the cla-already-signed label Sep 17, 2017

weitzhandler and others added 3 commits September 17, 2017 03:06

Merge branch 'master' into dataannotations-length-fix

5ace89b

Extracted code from try-catch.

3703fa6

Extracted code from try-catch

6d02660

danmoseley requested a review from a user September 17, 2017 01:56

ghost approved these changes Sep 18, 2017

View reviewed changes

karelz added the area-System.ComponentModel.DataAnnotations label Sep 18, 2017

karelz assigned weitzhandler, divega, ajcvickers and lajones Sep 18, 2017

danmoseley reviewed Sep 18, 2017

View reviewed changes

Shimmy Weitzhandler added 2 commits September 18, 2017 22:32

Added comment as requested.

6170e27

Typo correction

eb67130

danmoseley merged commit b4daeab into dotnet:master Sep 18, 2017

karelz added this to the 2.1.0 milestone Oct 11, 2017

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Data-annotations length fix #24101

Data-annotations length fix #24101

weitzhandler commented Sep 17, 2017 •

edited

Loading

weitzhandler commented Sep 17, 2017

danmoseley Sep 18, 2017

weitzhandler Sep 18, 2017

danmoseley commented Sep 18, 2017

weitzhandler commented Sep 18, 2017 •

edited

Loading

ajcvickers commented Sep 18, 2017

weitzhandler commented Sep 18, 2017 •

edited

Loading

danmoseley commented Sep 19, 2017

danmoseley commented Sep 19, 2017

danmoseley commented Sep 19, 2017

ghost commented Sep 19, 2017

danmoseley commented Sep 19, 2017

ajcvickers commented Sep 19, 2017

danmoseley commented Sep 19, 2017

weitzhandler commented Sep 19, 2017 •

edited

Loading

ajcvickers commented Sep 20, 2017

Data-annotations length fix #24101

Data-annotations length fix #24101

Conversation

weitzhandler commented Sep 17, 2017 • edited Loading

weitzhandler commented Sep 17, 2017

danmoseley Sep 18, 2017

Choose a reason for hiding this comment

weitzhandler Sep 18, 2017

Choose a reason for hiding this comment

danmoseley commented Sep 18, 2017

weitzhandler commented Sep 18, 2017 • edited Loading

ajcvickers commented Sep 18, 2017

weitzhandler commented Sep 18, 2017 • edited Loading

danmoseley commented Sep 19, 2017

danmoseley commented Sep 19, 2017

danmoseley commented Sep 19, 2017

ghost commented Sep 19, 2017

danmoseley commented Sep 19, 2017

ajcvickers commented Sep 19, 2017

danmoseley commented Sep 19, 2017

weitzhandler commented Sep 19, 2017 • edited Loading

ajcvickers commented Sep 20, 2017

weitzhandler commented Sep 17, 2017 •

edited

Loading

weitzhandler commented Sep 18, 2017 •

edited

Loading

weitzhandler commented Sep 18, 2017 •

edited

Loading

weitzhandler commented Sep 19, 2017 •

edited

Loading