-
Notifications
You must be signed in to change notification settings - Fork 4.8k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
don't run LongModuleFileNamesAreSupported test on x86 #57471
Conversation
Tagging subscribers to this area: @dotnet/area-system-diagnostics-process Issue Detailsfor some reason it's flaky fixes #57452
|
&& !PlatformDetection.IsMonoRuntime // Assembly.LoadFile used the way this test is implemented fails on Mono | ||
&& OperatingSystem.IsWindowsVersionAtLeast(8); // it's specific to Windows and does not work on Windows 7 | ||
&& PlatformDetection.Is64BitProcess; // for some reason it's flaky on x86 |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@JeremyKuhne do you perhaps know about any long path issues specific to x86?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
can you explain what do you mean by flaky? what's happening? I'd expect this to be deterministic
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
should this be !(Is32Bit && !IsArm)
- perhaps we should add IsX86 in the PlatformDetection - we should also open the issue to try to understand why this is failing on x86 I think (unless we don't care for some reason).
Alternative: RuntimeInformation.ProcessArchitecture != Architecture.X86
- you should still consider adding that to platform detection
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Perhaps we should add more diagnostics.. I.e. print all module list, enumerate modules in the assembly returned by Assembly.LoadFile. This sound rather sketchy as is - perhaps it's unrelated issue completely but we should at least understand why it's happening before we disable I think
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
RuntimeInformation.ProcessArchitecture != Architecture.X86
done
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@JeremyKuhne do you perhaps know about any long path issues specific to x86?
The only issues I've ever seen are bugs with the WOW layer when running an x86 binary on x64. I'm unaware of any outstanding issues though.
&& !PlatformDetection.IsMonoRuntime // Assembly.LoadFile used the way this test is implemented fails on Mono | ||
&& OperatingSystem.IsWindowsVersionAtLeast(8); // it's specific to Windows and does not work on Windows 7 | ||
&& RuntimeInformation.ProcessArchitecture != Architecture.X86; // for some reason it's flaky on x86 |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
We're confident this is an OS limitation rather than a product or test bug?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
My guess is that we end up taking a fallback path in the assembly loader on x86 for some reason. The assembly loader has number of fallback paths for historic reasons to compensate for OS and compiler issues.
The assembly loader behavior that this test depends on is not guaranteed. The test is disabled on Mono for this reason. If once we enable these tests in single-file or NativeAOT modes, the test may need to be disabled for those configurations as well.
It would be better if the test came with its own .dll and called LoadLibrary
on it. Then it can depend on that the library has to show up in the list.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@jkotas I've followed your suggestion and used LoadLibrary
directly. On one of the x86 CI legs it fails with ERROR_MOD_NOT_FOUND
:
System.Diagnostics.Tests.ProcessModuleTests.LongModuleFileNamesAreSupported [FAIL]
LoadLibrary failed with 126
Expected: False
Actual: True
Stack Trace:
/_/src/libraries/System.Diagnostics.Process/tests/ProcessModuleTests.Windows.cs(37,0): at System.Diagnostics.Tests.ProcessModuleTests.LongModuleFileNamesAreSupported()
I can't repro it locally and find out which module is missing (this is just a lib with a single class). Since this test is basically best effort to test long paths support for process module file paths, would it be OK to just skip the test on ERROR_MOD_NOT_FOUND
on x86?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yes, I think it is ok to skip the test when LoadLibrary fails to load the module with the long file name. My guess is that the long paths are not properly enabled on the system when it happens.
…$(CommonPath)Interop\ can be included in the test project
/azp run runtime |
Azure Pipelines successfully started running 1 pipeline(s). |
If this is the solution to #57452 is definitely 6.0 milestone |
{ | ||
internal static partial class Interop |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
this Interop
type was causing a conflict with the other Interop
that we use in product code and because of that I was not able to call Interop.Kernel32.LoadLibrary
in the test code. I decided to just remove the namespace so it de-facto becames a single Interop
type.
the CI failures are not related, as soon as I get an approval the PR is ready to be merged |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM
This particular tests uses
Assembly.LoadFile(path)
to load a test assembly:runtime/src/libraries/System.Diagnostics.Process/tests/ProcessModuleTests.cs
Lines 101 to 107 in 80392a9
for some reason it's flaky on x86. I've tried to repro it on two Windows machines and failed. Since I don't have better idea, I think it's just OK not to run this test on x86.
The previous check of Windows version (
OperatingSystem.IsWindowsVersionAtLeast(8)
) can be removed as in the past this test was failing only on Windows 7 x86.fixes #57452