[cdac] Start Loader contract and implement ISOSDacInterface::GetModuleData in cDAC

[elinor-fung]

• Start a Loader contract - currently contains what is needed for GetModuleData
• Implement ISOSDacInterface::GetModuleData in cDAC
• Store base address and is reflection emit bit on Module for easier diagnostics access

Contributes to #99302

[dotnet-policy-service]

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

[elinor-fung]

This was the only value I found in dotnet/diagnostics and microsoft/clrmd that was actually used. And it was just to print an explicit IS_EDIT_AND_CONTINUE as part of dumpmodule (also prints the raw flag values).

[lambdageek]

I wonder if we're going to have a problem in practice with the cDAC being more eager to read data from the target process up front. I think the brittle DAC is lazier about following pointers in target memory. It might be able to emit more info to the debugger before failing.

[davidwrighton]

This is a concern for mini and triage dumps only. What we need to do for those is make sure that our routines for ensuring that enough data structures are live are updated so that they make sure to cover the important scenarios needed for examining dumps. Also we will likely need to have the ability for some of the apis to produce incomplete results. The only reason this works in the existing scheme is that there are a selected set of places in the DAC where failures are just ignored. We will eventually need to add the same things to the cDAC, but I don't think we should do that yet.

[elinor-fung]

• Start a Loader contract - currently contains what is needed for GetModuleData
• Implement ISOSDacInterface::GetModuleData in cDAC
• Store base address and is reflection emit bit on Module for easier diagnostics access

Contributes to #99302

[dotnet-policy-service]

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

[elinor-fung]

This was the only value I found in dotnet/diagnostics and microsoft/clrmd that was actually used. And it was just to print an explicit IS_EDIT_AND_CONTINUE as part of dumpmodule (also prints the raw flag values).

[lambdageek]

I wonder if we're going to have a problem in practice with the cDAC being more eager to read data from the target process up front. I think the brittle DAC is lazier about following pointers in target memory. It might be able to emit more info to the debugger before failing.

[davidwrighton]

This is a concern for mini and triage dumps only. What we need to do for those is make sure that our routines for ensuring that enough data structures are live are updated so that they make sure to cover the important scenarios needed for examining dumps. Also we will likely need to have the ability for some of the apis to produce incomplete results. The only reason this works in the existing scheme is that there are a selected set of places in the DAC where failures are just ignored. We will eventually need to add the same things to the cDAC, but I don't think we should do that yet.

[elinor-fung]

• Start a Loader contract - currently contains what is needed for GetModuleData
• Implement ISOSDacInterface::GetModuleData in cDAC
• Store base address and is reflection emit bit on Module for easier diagnostics access

Contributes to #99302

[dotnet-policy-service]

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

[elinor-fung]

This was the only value I found in dotnet/diagnostics and microsoft/clrmd that was actually used. And it was just to print an explicit IS_EDIT_AND_CONTINUE as part of dumpmodule (also prints the raw flag values).

[lambdageek]

I wonder if we're going to have a problem in practice with the cDAC being more eager to read data from the target process up front. I think the brittle DAC is lazier about following pointers in target memory. It might be able to emit more info to the debugger before failing.

[davidwrighton]

This is a concern for mini and triage dumps only. What we need to do for those is make sure that our routines for ensuring that enough data structures are live are updated so that they make sure to cover the important scenarios needed for examining dumps. Also we will likely need to have the ability for some of the apis to produce incomplete results. The only reason this works in the existing scheme is that there are a selected set of places in the DAC where failures are just ignored. We will eventually need to add the same things to the cDAC, but I don't think we should do that yet.

[elinor-fung]

• Start a Loader contract - currently contains what is needed for GetModuleData
• Implement ISOSDacInterface::GetModuleData in cDAC
• Store base address and is reflection emit bit on Module for easier diagnostics access

Contributes to #99302

[dotnet-policy-service]

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

[elinor-fung]

This was the only value I found in dotnet/diagnostics and microsoft/clrmd that was actually used. And it was just to print an explicit IS_EDIT_AND_CONTINUE as part of dumpmodule (also prints the raw flag values).

[lambdageek]

I wonder if we're going to have a problem in practice with the cDAC being more eager to read data from the target process up front. I think the brittle DAC is lazier about following pointers in target memory. It might be able to emit more info to the debugger before failing.

[davidwrighton]

This is a concern for mini and triage dumps only. What we need to do for those is make sure that our routines for ensuring that enough data structures are live are updated so that they make sure to cover the important scenarios needed for examining dumps. Also we will likely need to have the ability for some of the apis to produce incomplete results. The only reason this works in the existing scheme is that there are a selected set of places in the DAC where failures are just ignored. We will eventually need to add the same things to the cDAC, but I don't think we should do that yet.

[elinor-fung]

• Start a Loader contract - currently contains what is needed for GetModuleData
• Implement ISOSDacInterface::GetModuleData in cDAC
• Store base address and is reflection emit bit on Module for easier diagnostics access

Contributes to #99302

[dotnet-policy-service]

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

[elinor-fung]

This was the only value I found in dotnet/diagnostics and microsoft/clrmd that was actually used. And it was just to print an explicit IS_EDIT_AND_CONTINUE as part of dumpmodule (also prints the raw flag values).

[lambdageek]

I wonder if we're going to have a problem in practice with the cDAC being more eager to read data from the target process up front. I think the brittle DAC is lazier about following pointers in target memory. It might be able to emit more info to the debugger before failing.

[davidwrighton]

This is a concern for mini and triage dumps only. What we need to do for those is make sure that our routines for ensuring that enough data structures are live are updated so that they make sure to cover the important scenarios needed for examining dumps. Also we will likely need to have the ability for some of the apis to produce incomplete results. The only reason this works in the existing scheme is that there are a selected set of places in the DAC where failures are just ignored. We will eventually need to add the same things to the cDAC, but I don't think we should do that yet.

[elinor-fung]

• Start a Loader contract - currently contains what is needed for GetModuleData
• Implement ISOSDacInterface::GetModuleData in cDAC
• Store base address and is reflection emit bit on Module for easier diagnostics access

Contributes to #99302

[dotnet-policy-service]

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

[elinor-fung]

This was the only value I found in dotnet/diagnostics and microsoft/clrmd that was actually used. And it was just to print an explicit IS_EDIT_AND_CONTINUE as part of dumpmodule (also prints the raw flag values).

[lambdageek]

I wonder if we're going to have a problem in practice with the cDAC being more eager to read data from the target process up front. I think the brittle DAC is lazier about following pointers in target memory. It might be able to emit more info to the debugger before failing.

[davidwrighton]

This is a concern for mini and triage dumps only. What we need to do for those is make sure that our routines for ensuring that enough data structures are live are updated so that they make sure to cover the important scenarios needed for examining dumps. Also we will likely need to have the ability for some of the apis to produce incomplete results. The only reason this works in the existing scheme is that there are a selected set of places in the DAC where failures are just ignored. We will eventually need to add the same things to the cDAC, but I don't think we should do that yet.

[elinor-fung]

• Start a Loader contract - currently contains what is needed for GetModuleData
• Implement ISOSDacInterface::GetModuleData in cDAC
• Store base address and is reflection emit bit on Module for easier diagnostics access

Contributes to #99302

[dotnet-policy-service]

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

[elinor-fung]

This was the only value I found in dotnet/diagnostics and microsoft/clrmd that was actually used. And it was just to print an explicit IS_EDIT_AND_CONTINUE as part of dumpmodule (also prints the raw flag values).

[lambdageek]

I wonder if we're going to have a problem in practice with the cDAC being more eager to read data from the target process up front. I think the brittle DAC is lazier about following pointers in target memory. It might be able to emit more info to the debugger before failing.

[davidwrighton]

This is a concern for mini and triage dumps only. What we need to do for those is make sure that our routines for ensuring that enough data structures are live are updated so that they make sure to cover the important scenarios needed for examining dumps. Also we will likely need to have the ability for some of the apis to produce incomplete results. The only reason this works in the existing scheme is that there are a selected set of places in the DAC where failures are just ignored. We will eventually need to add the same things to the cDAC, but I don't think we should do that yet.

[elinor-fung]

• Start a Loader contract - currently contains what is needed for GetModuleData
• Implement ISOSDacInterface::GetModuleData in cDAC
• Store base address and is reflection emit bit on Module for easier diagnostics access

Contributes to #99302

[dotnet-policy-service]

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

[elinor-fung]

This was the only value I found in dotnet/diagnostics and microsoft/clrmd that was actually used. And it was just to print an explicit IS_EDIT_AND_CONTINUE as part of dumpmodule (also prints the raw flag values).

[lambdageek]

I wonder if we're going to have a problem in practice with the cDAC being more eager to read data from the target process up front. I think the brittle DAC is lazier about following pointers in target memory. It might be able to emit more info to the debugger before failing.

[davidwrighton]

This is a concern for mini and triage dumps only. What we need to do for those is make sure that our routines for ensuring that enough data structures are live are updated so that they make sure to cover the important scenarios needed for examining dumps. Also we will likely need to have the ability for some of the apis to produce incomplete results. The only reason this works in the existing scheme is that there are a selected set of places in the DAC where failures are just ignored. We will eventually need to add the same things to the cDAC, but I don't think we should do that yet.