Cache the path on Module for easier diagnostics

[elinor-fung]

• Make PEImage take a path in its constructor and make its member const
• Store the path when initializing Module

I'm trying to avoid having to expose PEAssembly/PEImage in data descriptors to the cDAC just to get the module path.
Contributes to #99302

cc @lambdageek@davidwrighton

[dotnet-policy-service]

Tagging subscribers to this area: @mangod9
See info in area-owners.md if you want to be subscribed.

[jkotas]

Are we also trying to avoid exposing SString for the cdac?

SString works well for temporary values - method arguments, locals, etc. It works poorly as a field with longer lifetime that is potentially accessed by multiple threads (you have to be very careful how you use it to avoid subtle race conditions). I think it is a good idea to avoid exposing SString for the cdac

[AaronRobinsonMSFT]

potentially accessed by multiple threads

@lambdageek You would think marking something const, as @elinor-fung did in this PR, would help mitigate some of those things. The issue is the SString type was written very badly and the designers applied const to methods and then when in the member function would cast away const. It is infuriating and means that applying const to an SString instance really doesn't mean much. For example, calling const UTF8 *GetUTF8() const; on an instance could mutate the instance and replace the UTF-16 encoding with a UTF-8 one while another thread is using it. I'm hoping I can get to it in .NET 10, but I've been saying that since .NET 8 :-(

[elinor-fung]

Re: lifetime and const SString not really being const

• Module ref counts PEAssembly which ref counts PEImage which has the path, so the underlying string should not go away for the lifetime of the Module
• As @AaronRobinsonMSFT mentioned, making const actually mean const for SString is non-trivial. I just added an assert to Module::GetPath around the cached path matching the one in PEImage

[davidwrighton]

Looks good to me. Long term, I'd like to see us stop using an SString as a field at all on PEImage, but we can do that some other time.

[elinor-fung]

• Make PEImage take a path in its constructor and make its member const
• Store the path when initializing Module

I'm trying to avoid having to expose PEAssembly/PEImage in data descriptors to the cDAC just to get the module path.
Contributes to #99302

cc @lambdageek@davidwrighton

[dotnet-policy-service]

Tagging subscribers to this area: @mangod9
See info in area-owners.md if you want to be subscribed.

[jkotas]

Are we also trying to avoid exposing SString for the cdac?

SString works well for temporary values - method arguments, locals, etc. It works poorly as a field with longer lifetime that is potentially accessed by multiple threads (you have to be very careful how you use it to avoid subtle race conditions). I think it is a good idea to avoid exposing SString for the cdac

[AaronRobinsonMSFT]

potentially accessed by multiple threads

@lambdageek You would think marking something const, as @elinor-fung did in this PR, would help mitigate some of those things. The issue is the SString type was written very badly and the designers applied const to methods and then when in the member function would cast away const. It is infuriating and means that applying const to an SString instance really doesn't mean much. For example, calling const UTF8 *GetUTF8() const; on an instance could mutate the instance and replace the UTF-16 encoding with a UTF-8 one while another thread is using it. I'm hoping I can get to it in .NET 10, but I've been saying that since .NET 8 :-(

[elinor-fung]

Re: lifetime and const SString not really being const

• Module ref counts PEAssembly which ref counts PEImage which has the path, so the underlying string should not go away for the lifetime of the Module
• As @AaronRobinsonMSFT mentioned, making const actually mean const for SString is non-trivial. I just added an assert to Module::GetPath around the cached path matching the one in PEImage

[davidwrighton]

Looks good to me. Long term, I'd like to see us stop using an SString as a field at all on PEImage, but we can do that some other time.

[elinor-fung]

• Make PEImage take a path in its constructor and make its member const
• Store the path when initializing Module

I'm trying to avoid having to expose PEAssembly/PEImage in data descriptors to the cDAC just to get the module path.
Contributes to #99302

cc @lambdageek@davidwrighton

[dotnet-policy-service]

Tagging subscribers to this area: @mangod9
See info in area-owners.md if you want to be subscribed.

[jkotas]

Are we also trying to avoid exposing SString for the cdac?

SString works well for temporary values - method arguments, locals, etc. It works poorly as a field with longer lifetime that is potentially accessed by multiple threads (you have to be very careful how you use it to avoid subtle race conditions). I think it is a good idea to avoid exposing SString for the cdac

[AaronRobinsonMSFT]

potentially accessed by multiple threads

@lambdageek You would think marking something const, as @elinor-fung did in this PR, would help mitigate some of those things. The issue is the SString type was written very badly and the designers applied const to methods and then when in the member function would cast away const. It is infuriating and means that applying const to an SString instance really doesn't mean much. For example, calling const UTF8 *GetUTF8() const; on an instance could mutate the instance and replace the UTF-16 encoding with a UTF-8 one while another thread is using it. I'm hoping I can get to it in .NET 10, but I've been saying that since .NET 8 :-(

[elinor-fung]

Re: lifetime and const SString not really being const

• Module ref counts PEAssembly which ref counts PEImage which has the path, so the underlying string should not go away for the lifetime of the Module
• As @AaronRobinsonMSFT mentioned, making const actually mean const for SString is non-trivial. I just added an assert to Module::GetPath around the cached path matching the one in PEImage

[davidwrighton]

Looks good to me. Long term, I'd like to see us stop using an SString as a field at all on PEImage, but we can do that some other time.

[elinor-fung]

• Make PEImage take a path in its constructor and make its member const
• Store the path when initializing Module

I'm trying to avoid having to expose PEAssembly/PEImage in data descriptors to the cDAC just to get the module path.
Contributes to #99302

cc @lambdageek@davidwrighton

[dotnet-policy-service]

Tagging subscribers to this area: @mangod9
See info in area-owners.md if you want to be subscribed.

[jkotas]

Are we also trying to avoid exposing SString for the cdac?

SString works well for temporary values - method arguments, locals, etc. It works poorly as a field with longer lifetime that is potentially accessed by multiple threads (you have to be very careful how you use it to avoid subtle race conditions). I think it is a good idea to avoid exposing SString for the cdac

[AaronRobinsonMSFT]

potentially accessed by multiple threads

@lambdageek You would think marking something const, as @elinor-fung did in this PR, would help mitigate some of those things. The issue is the SString type was written very badly and the designers applied const to methods and then when in the member function would cast away const. It is infuriating and means that applying const to an SString instance really doesn't mean much. For example, calling const UTF8 *GetUTF8() const; on an instance could mutate the instance and replace the UTF-16 encoding with a UTF-8 one while another thread is using it. I'm hoping I can get to it in .NET 10, but I've been saying that since .NET 8 :-(

[elinor-fung]

Re: lifetime and const SString not really being const

• Module ref counts PEAssembly which ref counts PEImage which has the path, so the underlying string should not go away for the lifetime of the Module
• As @AaronRobinsonMSFT mentioned, making const actually mean const for SString is non-trivial. I just added an assert to Module::GetPath around the cached path matching the one in PEImage

[davidwrighton]

Looks good to me. Long term, I'd like to see us stop using an SString as a field at all on PEImage, but we can do that some other time.

[elinor-fung]

• Make PEImage take a path in its constructor and make its member const
• Store the path when initializing Module

I'm trying to avoid having to expose PEAssembly/PEImage in data descriptors to the cDAC just to get the module path.
Contributes to #99302

cc @lambdageek@davidwrighton

[dotnet-policy-service]

Tagging subscribers to this area: @mangod9
See info in area-owners.md if you want to be subscribed.

[jkotas]

Are we also trying to avoid exposing SString for the cdac?

SString works well for temporary values - method arguments, locals, etc. It works poorly as a field with longer lifetime that is potentially accessed by multiple threads (you have to be very careful how you use it to avoid subtle race conditions). I think it is a good idea to avoid exposing SString for the cdac

[AaronRobinsonMSFT]

potentially accessed by multiple threads

@lambdageek You would think marking something const, as @elinor-fung did in this PR, would help mitigate some of those things. The issue is the SString type was written very badly and the designers applied const to methods and then when in the member function would cast away const. It is infuriating and means that applying const to an SString instance really doesn't mean much. For example, calling const UTF8 *GetUTF8() const; on an instance could mutate the instance and replace the UTF-16 encoding with a UTF-8 one while another thread is using it. I'm hoping I can get to it in .NET 10, but I've been saying that since .NET 8 :-(

[elinor-fung]

Re: lifetime and const SString not really being const

• Module ref counts PEAssembly which ref counts PEImage which has the path, so the underlying string should not go away for the lifetime of the Module
• As @AaronRobinsonMSFT mentioned, making const actually mean const for SString is non-trivial. I just added an assert to Module::GetPath around the cached path matching the one in PEImage

[davidwrighton]

Looks good to me. Long term, I'd like to see us stop using an SString as a field at all on PEImage, but we can do that some other time.

[elinor-fung]

• Make PEImage take a path in its constructor and make its member const
• Store the path when initializing Module

I'm trying to avoid having to expose PEAssembly/PEImage in data descriptors to the cDAC just to get the module path.
Contributes to #99302

cc @lambdageek@davidwrighton

[dotnet-policy-service]

Tagging subscribers to this area: @mangod9
See info in area-owners.md if you want to be subscribed.

[jkotas]

Are we also trying to avoid exposing SString for the cdac?

SString works well for temporary values - method arguments, locals, etc. It works poorly as a field with longer lifetime that is potentially accessed by multiple threads (you have to be very careful how you use it to avoid subtle race conditions). I think it is a good idea to avoid exposing SString for the cdac

[AaronRobinsonMSFT]

potentially accessed by multiple threads

@lambdageek You would think marking something const, as @elinor-fung did in this PR, would help mitigate some of those things. The issue is the SString type was written very badly and the designers applied const to methods and then when in the member function would cast away const. It is infuriating and means that applying const to an SString instance really doesn't mean much. For example, calling const UTF8 *GetUTF8() const; on an instance could mutate the instance and replace the UTF-16 encoding with a UTF-8 one while another thread is using it. I'm hoping I can get to it in .NET 10, but I've been saying that since .NET 8 :-(

[elinor-fung]

Re: lifetime and const SString not really being const

• Module ref counts PEAssembly which ref counts PEImage which has the path, so the underlying string should not go away for the lifetime of the Module
• As @AaronRobinsonMSFT mentioned, making const actually mean const for SString is non-trivial. I just added an assert to Module::GetPath around the cached path matching the one in PEImage

[davidwrighton]

Looks good to me. Long term, I'd like to see us stop using an SString as a field at all on PEImage, but we can do that some other time.

[elinor-fung]

• Make PEImage take a path in its constructor and make its member const
• Store the path when initializing Module

I'm trying to avoid having to expose PEAssembly/PEImage in data descriptors to the cDAC just to get the module path.
Contributes to #99302

cc @lambdageek@davidwrighton

[dotnet-policy-service]

Tagging subscribers to this area: @mangod9
See info in area-owners.md if you want to be subscribed.

[jkotas]

Are we also trying to avoid exposing SString for the cdac?

SString works well for temporary values - method arguments, locals, etc. It works poorly as a field with longer lifetime that is potentially accessed by multiple threads (you have to be very careful how you use it to avoid subtle race conditions). I think it is a good idea to avoid exposing SString for the cdac

[AaronRobinsonMSFT]

potentially accessed by multiple threads

@lambdageek You would think marking something const, as @elinor-fung did in this PR, would help mitigate some of those things. The issue is the SString type was written very badly and the designers applied const to methods and then when in the member function would cast away const. It is infuriating and means that applying const to an SString instance really doesn't mean much. For example, calling const UTF8 *GetUTF8() const; on an instance could mutate the instance and replace the UTF-16 encoding with a UTF-8 one while another thread is using it. I'm hoping I can get to it in .NET 10, but I've been saying that since .NET 8 :-(

[elinor-fung]

Re: lifetime and const SString not really being const

• Module ref counts PEAssembly which ref counts PEImage which has the path, so the underlying string should not go away for the lifetime of the Module
• As @AaronRobinsonMSFT mentioned, making const actually mean const for SString is non-trivial. I just added an assert to Module::GetPath around the cached path matching the one in PEImage

[davidwrighton]

Looks good to me. Long term, I'd like to see us stop using an SString as a field at all on PEImage, but we can do that some other time.

[elinor-fung]

• Make PEImage take a path in its constructor and make its member const
• Store the path when initializing Module

I'm trying to avoid having to expose PEAssembly/PEImage in data descriptors to the cDAC just to get the module path.
Contributes to #99302

cc @lambdageek@davidwrighton

[dotnet-policy-service]

Tagging subscribers to this area: @mangod9
See info in area-owners.md if you want to be subscribed.

[jkotas]

Are we also trying to avoid exposing SString for the cdac?

SString works well for temporary values - method arguments, locals, etc. It works poorly as a field with longer lifetime that is potentially accessed by multiple threads (you have to be very careful how you use it to avoid subtle race conditions). I think it is a good idea to avoid exposing SString for the cdac

[AaronRobinsonMSFT]

potentially accessed by multiple threads

@lambdageek You would think marking something const, as @elinor-fung did in this PR, would help mitigate some of those things. The issue is the SString type was written very badly and the designers applied const to methods and then when in the member function would cast away const. It is infuriating and means that applying const to an SString instance really doesn't mean much. For example, calling const UTF8 *GetUTF8() const; on an instance could mutate the instance and replace the UTF-16 encoding with a UTF-8 one while another thread is using it. I'm hoping I can get to it in .NET 10, but I've been saying that since .NET 8 :-(

[elinor-fung]

Re: lifetime and const SString not really being const

• Module ref counts PEAssembly which ref counts PEImage which has the path, so the underlying string should not go away for the lifetime of the Module
• As @AaronRobinsonMSFT mentioned, making const actually mean const for SString is non-trivial. I just added an assert to Module::GetPath around the cached path matching the one in PEImage

[davidwrighton]

Looks good to me. Long term, I'd like to see us stop using an SString as a field at all on PEImage, but we can do that some other time.