Use size_t for hostfxr info count members

[am11]

Installer build with GCC failed:

  /runtime/src/native/corehost/fxr/hostfxr.cpp:456:35: error: narrowing conversion of 'environment_sdk_infos.std::vector<hostfxr_dotnet_environment_sdk_info>::size()' from 'std::vector<hostfxr_dotnet_environment_sdk_info>::size_type' {aka 'long unsigned int'} to 'int32_t' {aka 'int'} inside { } [-Werror=narrowing]
           environment_sdk_infos.size(),
           ~~~~~~~~~~~~~~~~~~~~~~~~~~^~
  /runtime/src/native/corehost/fxr/hostfxr.cpp:458:41: error: narrowing conversion of 'environment_framework_infos.std::vector<hostfxr_dotnet_environment_framework_info>::size()' from 'std::vector<hostfxr_dotnet_environment_framework_info>::size_type' {aka 'long unsigned int'} to 'int32_t' {aka 'int'} inside { } [-Werror=narrowing]
           environment_framework_infos.size(),
           ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^~

This PR fixes the build.

[ghost]

Tagging subscribers to this area: @vitek-karas, @agocke
See info in area-owners.md if you want to be subscribed.

Issue Details

---

Installer build with GCC failed:

  /runtime/src/native/corehost/fxr/hostfxr.cpp:456:35: error: narrowing conversion of 'environment_sdk_infos.std::vector<hostfxr_dotnet_environment_sdk_info>::size()' from 'std::vector<hostfxr_dotnet_environment_sdk_info>::size_type' {aka 'long unsigned int'} to 'int32_t' {aka 'int'} inside { } [-Werror=narrowing]
           environment_sdk_infos.size(),
           ~~~~~~~~~~~~~~~~~~~~~~~~~~^~
  /runtime/src/native/corehost/fxr/hostfxr.cpp:458:41: error: narrowing conversion of 'environment_framework_infos.std::vector<hostfxr_dotnet_environment_framework_info>::size()' from 'std::vector<hostfxr_dotnet_environment_framework_info>::size_type' {aka 'long unsigned int'} to 'int32_t' {aka 'int'} inside { } [-Werror=narrowing]
           environment_framework_infos.size(),
           ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^~

This PR fixes the build.

Author:	am11
Assignees:	-
Labels:	area-Host
Milestone:	-

[am11]

cc @VSadov

[vitek-karas]

How come this made it through CI - is the warning disabled, or something like that?

[am11]

The gcc leg only builds CoreCLR subset using gcc v9. Build breakage was caught yesterday by downstream CI set up in the fork. With the fix, it passed.

I think Arch Linux and few distros are also using gcc to build the .NET runtime package. Currently all native components (including mono) build just fine with gcc, so IMO expanding the gcc CI coverage to build all components would be great.

[VSadov]

I think some warnings are disabled, or maybe we never build host.native with GCC.
I see these warnings as errors in #48254 though.

[VSadov]

There are lots of other warnings as well, but those do not seem to be warn-as-error kind.

https://dev.azure.com/dnceng/9ee6d478-d288-47f7-aacc-f6e6d082ae6d/_apis/build/builds/994542/logs/2088

[am11]

The warnings configs are mostly shared between coreclr and installer, our gcc build validation leg is set up to only build coreclr subset.

[am11]

other warnings

the most significant one is:

warning: 'emitter::emitAddrMode::amScale' is too small to hold all values of 'enum emitter::opSize'

having 1000+ occurrences. It has no warning code as it is emerging from core compiler which just writes to stderr without failing.

The appropriate warning code is -Woverflow with gcc and -Wbitfield-enum-conversion with clang, when we assign a value to the instance of enum that exceeds its own bitfield width. In this case, we have a value in enum which exceeds its own bitfild's width, but we are not assigning it to any instance (yet). So compiler thinks it's a matter of time and warns.. tracked by #33541 (comment).

[VSadov]

Thanks!!!