Add support for the large-offset stresslog format.

[jkoritzinsky]

Add support for the large-offset stresslog format defined in dotnet/runtime#83855. I've validated that the dumplog command supports both the V<=3 format as well as the large-offset format with this change.

[mikem8361]

It looks good as long as the dumplog command still works for older version 6 and 7 runtimes. There isn't a lot (or any) testing of dumplog in the CI jobs so hopefully you can test it manually across all the supported runtimes/architectures.

[jkoritzinsky]

I'll do some additional testing with the older version runtimes (I've validated against main with these changes, but not old runtime versions) to validate the behavior.

[jkoritzinsky]

I've validated that this works downlevel for older target runtimes. Marking as no-merge for now until I get approval on the runtime side of this change.

[leculver]

Is it worth considering moving stresslogs to be a defined Dac API instead of SOS natively understanding the format? We'd have less "sos breaking changes" if we defined a stable API, and if the format changes so much we could add another interface, just like other features.

In the big long list of things that we still take a hard dependency on, I think it's just down to:

• StressLogs
• GCInfo
• GCDesc (?)

Or is there a reason we treat StressLogs differently from everything else?

(This would be for future consideration, or just my own edification. No need to block this PR or feature on this question.)

[mikem8361]

Is it worth considering moving stresslogs to be a defined Dac API instead of SOS natively understanding the format? We'd have less "sos breaking changes" if we defined a stable API, and if the format changes so much we could add another interface, just like other features.

In the big long list of things that we still take a hard dependency on, I think it's just down to:

• StressLogs
• GCInfo
• GCDesc (?)

Or is there a reason we treat StressLogs differently from everything else?

(This would be for future consideration, or just my own edification. No need to block this PR or feature on this question.)

I just noticed Lee's question was never questioned: It was just historical. I think at some point when SOS was part of the runtime before I moved it to it's own repo having these dependencies (I think the only other dependencies are the on TLS slots on Windows and the ThreadState bits) was easier and whomever added the dumplog command was a little lazy (I'm just guessing) and didn't go through the DAC.

[jkoritzinsky]

Removing DO NOT MERGE label as the dotnet/runtime PR is also approved will be merged soon.

[jkoritzinsky]

Add support for the large-offset stresslog format defined in dotnet/runtime#83855. I've validated that the dumplog command supports both the V<=3 format as well as the large-offset format with this change.

[mikem8361]

It looks good as long as the dumplog command still works for older version 6 and 7 runtimes. There isn't a lot (or any) testing of dumplog in the CI jobs so hopefully you can test it manually across all the supported runtimes/architectures.

[jkoritzinsky]

I'll do some additional testing with the older version runtimes (I've validated against main with these changes, but not old runtime versions) to validate the behavior.

[jkoritzinsky]

I've validated that this works downlevel for older target runtimes. Marking as no-merge for now until I get approval on the runtime side of this change.

[leculver]

Is it worth considering moving stresslogs to be a defined Dac API instead of SOS natively understanding the format? We'd have less "sos breaking changes" if we defined a stable API, and if the format changes so much we could add another interface, just like other features.

In the big long list of things that we still take a hard dependency on, I think it's just down to:

• StressLogs
• GCInfo
• GCDesc (?)

Or is there a reason we treat StressLogs differently from everything else?

(This would be for future consideration, or just my own edification. No need to block this PR or feature on this question.)

[mikem8361]

Is it worth considering moving stresslogs to be a defined Dac API instead of SOS natively understanding the format? We'd have less "sos breaking changes" if we defined a stable API, and if the format changes so much we could add another interface, just like other features.

In the big long list of things that we still take a hard dependency on, I think it's just down to:

• StressLogs
• GCInfo
• GCDesc (?)

Or is there a reason we treat StressLogs differently from everything else?

(This would be for future consideration, or just my own edification. No need to block this PR or feature on this question.)

I just noticed Lee's question was never questioned: It was just historical. I think at some point when SOS was part of the runtime before I moved it to it's own repo having these dependencies (I think the only other dependencies are the on TLS slots on Windows and the ThreadState bits) was easier and whomever added the dumplog command was a little lazy (I'm just guessing) and didn't go through the DAC.

[jkoritzinsky]

Removing DO NOT MERGE label as the dotnet/runtime PR is also approved will be merged soon.

[jkoritzinsky]

Add support for the large-offset stresslog format defined in dotnet/runtime#83855. I've validated that the dumplog command supports both the V<=3 format as well as the large-offset format with this change.

[mikem8361]

It looks good as long as the dumplog command still works for older version 6 and 7 runtimes. There isn't a lot (or any) testing of dumplog in the CI jobs so hopefully you can test it manually across all the supported runtimes/architectures.

[jkoritzinsky]

I'll do some additional testing with the older version runtimes (I've validated against main with these changes, but not old runtime versions) to validate the behavior.

[jkoritzinsky]

I've validated that this works downlevel for older target runtimes. Marking as no-merge for now until I get approval on the runtime side of this change.

[leculver]

Is it worth considering moving stresslogs to be a defined Dac API instead of SOS natively understanding the format? We'd have less "sos breaking changes" if we defined a stable API, and if the format changes so much we could add another interface, just like other features.

In the big long list of things that we still take a hard dependency on, I think it's just down to:

• StressLogs
• GCInfo
• GCDesc (?)

Or is there a reason we treat StressLogs differently from everything else?

(This would be for future consideration, or just my own edification. No need to block this PR or feature on this question.)

[mikem8361]

Is it worth considering moving stresslogs to be a defined Dac API instead of SOS natively understanding the format? We'd have less "sos breaking changes" if we defined a stable API, and if the format changes so much we could add another interface, just like other features.

In the big long list of things that we still take a hard dependency on, I think it's just down to:

• StressLogs
• GCInfo
• GCDesc (?)

Or is there a reason we treat StressLogs differently from everything else?

(This would be for future consideration, or just my own edification. No need to block this PR or feature on this question.)

I just noticed Lee's question was never questioned: It was just historical. I think at some point when SOS was part of the runtime before I moved it to it's own repo having these dependencies (I think the only other dependencies are the on TLS slots on Windows and the ThreadState bits) was easier and whomever added the dumplog command was a little lazy (I'm just guessing) and didn't go through the DAC.

[jkoritzinsky]

Removing DO NOT MERGE label as the dotnet/runtime PR is also approved will be merged soon.

[jkoritzinsky]

Add support for the large-offset stresslog format defined in dotnet/runtime#83855. I've validated that the dumplog command supports both the V<=3 format as well as the large-offset format with this change.

[mikem8361]

It looks good as long as the dumplog command still works for older version 6 and 7 runtimes. There isn't a lot (or any) testing of dumplog in the CI jobs so hopefully you can test it manually across all the supported runtimes/architectures.

[jkoritzinsky]

I'll do some additional testing with the older version runtimes (I've validated against main with these changes, but not old runtime versions) to validate the behavior.

[jkoritzinsky]

I've validated that this works downlevel for older target runtimes. Marking as no-merge for now until I get approval on the runtime side of this change.

[leculver]

Is it worth considering moving stresslogs to be a defined Dac API instead of SOS natively understanding the format? We'd have less "sos breaking changes" if we defined a stable API, and if the format changes so much we could add another interface, just like other features.

In the big long list of things that we still take a hard dependency on, I think it's just down to:

• StressLogs
• GCInfo
• GCDesc (?)

Or is there a reason we treat StressLogs differently from everything else?

(This would be for future consideration, or just my own edification. No need to block this PR or feature on this question.)

[mikem8361]

Is it worth considering moving stresslogs to be a defined Dac API instead of SOS natively understanding the format? We'd have less "sos breaking changes" if we defined a stable API, and if the format changes so much we could add another interface, just like other features.

In the big long list of things that we still take a hard dependency on, I think it's just down to:

• StressLogs
• GCInfo
• GCDesc (?)

Or is there a reason we treat StressLogs differently from everything else?

(This would be for future consideration, or just my own edification. No need to block this PR or feature on this question.)

I just noticed Lee's question was never questioned: It was just historical. I think at some point when SOS was part of the runtime before I moved it to it's own repo having these dependencies (I think the only other dependencies are the on TLS slots on Windows and the ThreadState bits) was easier and whomever added the dumplog command was a little lazy (I'm just guessing) and didn't go through the DAC.

[jkoritzinsky]

Removing DO NOT MERGE label as the dotnet/runtime PR is also approved will be merged soon.

[jkoritzinsky]

Add support for the large-offset stresslog format defined in dotnet/runtime#83855. I've validated that the dumplog command supports both the V<=3 format as well as the large-offset format with this change.

[mikem8361]

It looks good as long as the dumplog command still works for older version 6 and 7 runtimes. There isn't a lot (or any) testing of dumplog in the CI jobs so hopefully you can test it manually across all the supported runtimes/architectures.

[jkoritzinsky]

I'll do some additional testing with the older version runtimes (I've validated against main with these changes, but not old runtime versions) to validate the behavior.

[jkoritzinsky]

I've validated that this works downlevel for older target runtimes. Marking as no-merge for now until I get approval on the runtime side of this change.

[leculver]

Is it worth considering moving stresslogs to be a defined Dac API instead of SOS natively understanding the format? We'd have less "sos breaking changes" if we defined a stable API, and if the format changes so much we could add another interface, just like other features.

In the big long list of things that we still take a hard dependency on, I think it's just down to:

• StressLogs
• GCInfo
• GCDesc (?)

Or is there a reason we treat StressLogs differently from everything else?

(This would be for future consideration, or just my own edification. No need to block this PR or feature on this question.)

[mikem8361]

Is it worth considering moving stresslogs to be a defined Dac API instead of SOS natively understanding the format? We'd have less "sos breaking changes" if we defined a stable API, and if the format changes so much we could add another interface, just like other features.

In the big long list of things that we still take a hard dependency on, I think it's just down to:

• StressLogs
• GCInfo
• GCDesc (?)

Or is there a reason we treat StressLogs differently from everything else?

(This would be for future consideration, or just my own edification. No need to block this PR or feature on this question.)

I just noticed Lee's question was never questioned: It was just historical. I think at some point when SOS was part of the runtime before I moved it to it's own repo having these dependencies (I think the only other dependencies are the on TLS slots on Windows and the ThreadState bits) was easier and whomever added the dumplog command was a little lazy (I'm just guessing) and didn't go through the DAC.

[jkoritzinsky]

Removing DO NOT MERGE label as the dotnet/runtime PR is also approved will be merged soon.

[jkoritzinsky]

Add support for the large-offset stresslog format defined in dotnet/runtime#83855. I've validated that the dumplog command supports both the V<=3 format as well as the large-offset format with this change.

[mikem8361]

It looks good as long as the dumplog command still works for older version 6 and 7 runtimes. There isn't a lot (or any) testing of dumplog in the CI jobs so hopefully you can test it manually across all the supported runtimes/architectures.

[jkoritzinsky]

I'll do some additional testing with the older version runtimes (I've validated against main with these changes, but not old runtime versions) to validate the behavior.

[jkoritzinsky]

I've validated that this works downlevel for older target runtimes. Marking as no-merge for now until I get approval on the runtime side of this change.

[leculver]

Is it worth considering moving stresslogs to be a defined Dac API instead of SOS natively understanding the format? We'd have less "sos breaking changes" if we defined a stable API, and if the format changes so much we could add another interface, just like other features.

In the big long list of things that we still take a hard dependency on, I think it's just down to:

• StressLogs
• GCInfo
• GCDesc (?)

Or is there a reason we treat StressLogs differently from everything else?

(This would be for future consideration, or just my own edification. No need to block this PR or feature on this question.)

[mikem8361]

Is it worth considering moving stresslogs to be a defined Dac API instead of SOS natively understanding the format? We'd have less "sos breaking changes" if we defined a stable API, and if the format changes so much we could add another interface, just like other features.

In the big long list of things that we still take a hard dependency on, I think it's just down to:

• StressLogs
• GCInfo
• GCDesc (?)

Or is there a reason we treat StressLogs differently from everything else?

(This would be for future consideration, or just my own edification. No need to block this PR or feature on this question.)

I just noticed Lee's question was never questioned: It was just historical. I think at some point when SOS was part of the runtime before I moved it to it's own repo having these dependencies (I think the only other dependencies are the on TLS slots on Windows and the ThreadState bits) was easier and whomever added the dumplog command was a little lazy (I'm just guessing) and didn't go through the DAC.

[jkoritzinsky]

Removing DO NOT MERGE label as the dotnet/runtime PR is also approved will be merged soon.

[jkoritzinsky]

Add support for the large-offset stresslog format defined in dotnet/runtime#83855. I've validated that the dumplog command supports both the V<=3 format as well as the large-offset format with this change.

[mikem8361]

It looks good as long as the dumplog command still works for older version 6 and 7 runtimes. There isn't a lot (or any) testing of dumplog in the CI jobs so hopefully you can test it manually across all the supported runtimes/architectures.

[jkoritzinsky]

I'll do some additional testing with the older version runtimes (I've validated against main with these changes, but not old runtime versions) to validate the behavior.

[jkoritzinsky]

I've validated that this works downlevel for older target runtimes. Marking as no-merge for now until I get approval on the runtime side of this change.

[leculver]

Is it worth considering moving stresslogs to be a defined Dac API instead of SOS natively understanding the format? We'd have less "sos breaking changes" if we defined a stable API, and if the format changes so much we could add another interface, just like other features.

In the big long list of things that we still take a hard dependency on, I think it's just down to:

• StressLogs
• GCInfo
• GCDesc (?)

Or is there a reason we treat StressLogs differently from everything else?

(This would be for future consideration, or just my own edification. No need to block this PR or feature on this question.)

[mikem8361]

Is it worth considering moving stresslogs to be a defined Dac API instead of SOS natively understanding the format? We'd have less "sos breaking changes" if we defined a stable API, and if the format changes so much we could add another interface, just like other features.

In the big long list of things that we still take a hard dependency on, I think it's just down to:

• StressLogs
• GCInfo
• GCDesc (?)

Or is there a reason we treat StressLogs differently from everything else?

(This would be for future consideration, or just my own edification. No need to block this PR or feature on this question.)

I just noticed Lee's question was never questioned: It was just historical. I think at some point when SOS was part of the runtime before I moved it to it's own repo having these dependencies (I think the only other dependencies are the on TLS slots on Windows and the ThreadState bits) was easier and whomever added the dumplog command was a little lazy (I'm just guessing) and didn't go through the DAC.

[jkoritzinsky]

Removing DO NOT MERGE label as the dotnet/runtime PR is also approved will be merged soon.

[jkoritzinsky]

Add support for the large-offset stresslog format defined in dotnet/runtime#83855. I've validated that the dumplog command supports both the V<=3 format as well as the large-offset format with this change.

[mikem8361]

It looks good as long as the dumplog command still works for older version 6 and 7 runtimes. There isn't a lot (or any) testing of dumplog in the CI jobs so hopefully you can test it manually across all the supported runtimes/architectures.

[jkoritzinsky]

I'll do some additional testing with the older version runtimes (I've validated against main with these changes, but not old runtime versions) to validate the behavior.

[jkoritzinsky]

I've validated that this works downlevel for older target runtimes. Marking as no-merge for now until I get approval on the runtime side of this change.

[leculver]

Is it worth considering moving stresslogs to be a defined Dac API instead of SOS natively understanding the format? We'd have less "sos breaking changes" if we defined a stable API, and if the format changes so much we could add another interface, just like other features.

In the big long list of things that we still take a hard dependency on, I think it's just down to:

• StressLogs
• GCInfo
• GCDesc (?)

Or is there a reason we treat StressLogs differently from everything else?

(This would be for future consideration, or just my own edification. No need to block this PR or feature on this question.)

[mikem8361]

Is it worth considering moving stresslogs to be a defined Dac API instead of SOS natively understanding the format? We'd have less "sos breaking changes" if we defined a stable API, and if the format changes so much we could add another interface, just like other features.

In the big long list of things that we still take a hard dependency on, I think it's just down to:

• StressLogs
• GCInfo
• GCDesc (?)

Or is there a reason we treat StressLogs differently from everything else?

(This would be for future consideration, or just my own edification. No need to block this PR or feature on this question.)

I just noticed Lee's question was never questioned: It was just historical. I think at some point when SOS was part of the runtime before I moved it to it's own repo having these dependencies (I think the only other dependencies are the on TLS slots on Windows and the ThreadState bits) was easier and whomever added the dumplog command was a little lazy (I'm just guessing) and didn't go through the DAC.

[jkoritzinsky]

Removing DO NOT MERGE label as the dotnet/runtime PR is also approved will be merged soon.