search

dotnet / runtime

.NET is a cross-platform runtime for cloud, mobile, desktop, and IoT apps.
https://docs.microsoft.com/dotnet/core/
MIT License
14.95k stars 4.65k forks source link

Getting "Loading profiler failed during CoCreateInstance" error building on Alpine #66000

Open rakista112 opened 2 years ago

rakista112 commented 2 years ago

Description

We're getting a CoCreateInstance error when trying to load the profiler in my NetCore 3.0 app.

Reproduction Steps

  1. Build coreclr in docker with FROM mcr.microsoft.com/dotnet-buildtools/prereqs:alpine-3.15-20211214164127-f0ea7ba as base image

  2. use the release from https://codeload.github.com/dotnet/coreclr/tar.gz/release/3.0

  3. add these statements in the docker file
    RUN cd /code/coreclr/src/pal/src/libunwind/src/mi/ && sed -i '28d' dyn-info-list.c RUN cd /code/coreclr && ./build.sh x64 release configureonly skipgenerateversion nopgooptimize skiptests RUN cd /code/coreclr/bin/obj/Linux.x64.Release/src/pal && make

  4. Set CORECLR_ENABLE_PROFILING, CORECLR_PROFILER and CORECLR_PROFILER_PATH environment variables

  5. build your profiler with that library.

  6. Run the net core 3.0 application

Expected behavior

The profiler should attach and create logs

Actual behavior

This error is printed when the application runs. Loading profiler failed during CoCreateInstance. Profiler CLSID: '{************}'. HRESULT: 0x8007007e. Process ID (decimal): 161. Message ID: [0x2504]..NET Core SDK (3.1.416)

Regression?

No response

Known Workarounds

No response

Configuration

.net version: 3.1.416 Alpine x86_64 I don't know

Other information

No response

ghost commented 2 years ago

Tagging subscribers to this area: @tommcdon See info in area-owners.md if you want to be subscribed.

Issue Details
### Description We're getting a CoCreateInstance error when trying to load the profiler in my NetCore 3.0 app. ### Reproduction Steps 1. Build coreclr in docker with `FROM mcr.microsoft.com/dotnet-buildtools/prereqs:alpine-3.15-20211214164127-f0ea7ba` as base image 2. use the release from https://codeload.github.com/dotnet/coreclr/tar.gz/release/3.0 3. add these statements in the docker file `RUN cd /code/coreclr/src/pal/src/libunwind/src/mi/ && sed -i '28d' dyn-info-list.c` `RUN cd /code/coreclr && ./build.sh x64 release configureonly skipgenerateversion nopgooptimize skiptests` `RUN cd /code/coreclr/bin/obj/Linux.x64.Release/src/pal && make` 4. Set CORECLR_ENABLE_PROFILING, CORECLR_PROFILER and CORECLR_PROFILER_PATH environment variables 5. build your profiler with that library. 6. Run the net core 3.0 application ### Expected behavior The profiler should attach and create logs ### Actual behavior This error is printed when the application runs. `Loading profiler failed during CoCreateInstance. Profiler CLSID: '{************}'. HRESULT: 0x8007007e. Process ID (decimal): 161. Message ID: [0x2504]..NET Core SDK (3.1.416)` ### Regression? _No response_ ### Known Workarounds _No response_ ### Configuration .net version: 3.1.416 Alpine x86_64 I don't know ### Other information _No response_
Author: rakista112
Assignees: -
Labels: `area-Diagnostics-coreclr`, `untriaged`
Milestone: -
rakista112 commented 2 years ago

When I run it with strace, I see these errors before it writes the cocreate instance error open("/lib/libuuid.so.1", O_RDONLY|O_LARGEFILE|O_CLOEXEC) = -1 ENOENT (No such file or directory) open("/usr/local/lib/libuuid.so.1", O_RDONLY|O_LARGEFILE|O_CLOEXEC) = -1 ENOENT (No such file or directory) open("/usr/lib/libuuid.so.1", O_RDONLY|O_LARGEFILE|O_CLOEXEC) = -1 ENOENT (No such file or directory) open("/usr/share/locale/en_US.UTF-8/LC_MESSAGES/messages.mo", O_RDONLY|O_LARGEFILE) = -1 ENOENT (No such file or directory) open("/usr/share/locale/en_US.utf8/LC_MESSAGES/messages.mo", O_RDONLY|O_LARGEFILE) = -1 ENOENT (No such file or directory) open("/usr/share/locale/en_US/LC_MESSAGES/messages.mo", O_RDONLY|O_LARGEFILE) = -1 ENOENT (No such file or directory) open("/usr/share/locale/en.UTF-8/LC_MESSAGES/messages.mo", O_RDONLY|O_LARGEFILE) = -1 ENOENT (No such file or directory) open("/usr/share/locale/en.utf8/LC_MESSAGES/messages.mo", O_RDONLY|O_LARGEFILE) = -1 ENOENT (No such file or directory) open("/usr/share/locale/en/LC_MESSAGES/messages.mo", O_RDONLY|O_LARGEFILE) = -1 ENOENT (No such file or directory)

I fixed the uuid ones by running apk add libuuid but I don't know where to get the messages.mo files

tommcdon commented 2 years ago

@davmason

davmason commented 2 years ago

Hi @rakista112,

I am not familiar with that particular messages.mo file. This type of error is usually caused by your profiler having a dependency at build time that is unsatisfied at run time. It may be worthwhile to check your profiler build system to see what libraries you are building against, and making sure that all of those are installed in the container.

From searching around, it might be gettext or msgfmt, but I can't find anything definitive.

rakista112 commented 2 years ago

@davmason I tried installing gettext and still got the same messages. here's the full strace logs of just running an application strace -o strace_logs.txt dotnet AspNetCoreWebApp.dll, if it helps strace_logs.txt

davmason commented 2 years ago

Hi again @rakista112,

I am not sure how to help effectively here. The messages.mo is not provided as part of the runtime, it is something that is linked during the build process for the profiler you are trying to load.

You either have to find the library that provides it and install it on the target system, or statically link dependencies for the profiler

rakista112 commented 2 years ago

Okay, thanks I'll check my libraries/dependecies. I have a dependency on Boost 1.76 and coreclr PAL.

danmoseley commented 2 years ago

Just curious, why does it say CoCreateInstance when there is no COM on this platform - should that message be changed?

tommcdon commented 2 years ago

The hresult 0x8007007e is ERROR_MOD_NOT_FOUND - The specified module could not be found. And the error "Loading profiler failed during CoCreateInstance.." is IDS_E_PROF_CCI_FAILED: https://github.com/dotnet/runtime/blob/135e566b1fd4712c57f62e89234fca6fb2e5d674/src/coreclr/dlls/mscorrc/mscorrc.rc#L732

There error seems to be reported here: https://github.com/dotnet/runtime/blob/abb7df59eb4f8c813c07ddd6f46f94e59604577d/src/coreclr/vm/eetoprofinterfaceimpl.cpp#L321

Hope this helps!

danmoseley commented 2 years ago

On non-Windows, we still use a CLSID to instantiate the profiler, but it's an opaque GUID, and we handle it all ourselves? If that's the case, should we call these something different than a CLSID?

tommcdon commented 2 years ago

Just curious, why does it say CoCreateInstance when there is no COM on this platform - should that message be changed? On non-Windows, we still use a CLSID to instantiate the profiler, but it's an opaque GUID, and we handle it all ourselves? If that's the case, should we call these something different than a CLSID?

I agree the message is confusing for x-plat scenarios. It probably should say "Loading profiler failed" and instead of providing a CLSID it should provide the path.

The runtime implements a function called "FakeCoCreateInstance" which provides a cross-plat friendly mechanism for the profiler: https://github.com/dotnet/runtime/blob/d94ca5706c95b950719155905ec9a2d889160cbc/src/coreclr/utilcode/util.cpp#L233-L273

tommcdon commented 2 years ago

On non-Windows, we still use a CLSID to instantiate the profiler, but it's an opaque GUID, and we handle it all ourselves?

I believe we still use the CLSID in profiler class factory's CreateInstance call-