Resolve SigComparer regression introduced in 4.0

[ElektroKill]

Due to the urgency of this issue and the need for a quick fix, I have opened another pull request as opposed to #524 to demonstrate an alternative fix.

Advantages of this PR over #524:

• Makes the SigComparer in the next release backward compatible with the previous 3.6 one.
• Fixed the issue by making reference comparisons opt-in.
• Has clearly defined behavior in the flags and does not make SigComparer flags depend on other flags which has the benefits of clearer behavior and less chance of potentially introducing bugs later down the line.
• Adds convenience properties for reference comparing member definitions.

If this PR is merged, we can close #524 and quickly release a hotfix version for 4.0.

@wtfsck @CreateAndInject

[wtfsck]

@CreateAndInject Let me know if this fixes your issue.

[CreateAndInject]

Of course, CompareReferenceOnly isn't enbaled at all in this PR by default.
The only difference is my PR is radical, other program can use CompareReferenceOnly without any changes.
This PR is conservative, people must edit their code to use it explicitly.
E.g. Used By in dnSpy :
https://github.com/dnSpyEx/dnSpy/blob/e08de1db25d57766aa81ee452811ebf9d99c7d6a/Extensions/dnSpy.Analyzer/TreeNodes/Helpers.cs#L88-L89

[ElektroKill]

Of course, CompareReferenceOnly isn't enbaled at all in this PR by default. The only difference is my PR is radical, other program can use CompareReferenceOnly without any changes. This PR is conservative, people must edit their code to use it explicitly. E.g. Used By in dnSpy : dnSpyEx/dnSpy@e08de1d/Extensions/dnSpy.Analyzer/TreeNodes/Helpers.cs#L88-L89

Yes, the changes are designed to be conservative and require user to opt-in to this new behavior manually as to minimize issues when updating to the new version :p

[wtfsck]

Since it fixed the issue, I'm merging this one.

[ElektroKill]

When can we expect the hotfix version to be published to NuGet?

[wtfsck]

I just bumped the version, should take a few minutes for GitHub Actions to build and publish it.

[ElektroKill]

I just bumped the version, should take a few minutes for GitHub Actions to build and publish it.

It seems that for some reason the action did not execute the NuGet publish task 🤔

https://github.com/0xd4d/dnlib/actions/runs/6057961552/job/16439643254

[wtfsck]

Yeah I saw that, seems like a GitHub bug... On the main page, 4.1 is the latest release, when I go to releases, 4.0 is the latest release.

[ElektroKill]

Yeah I saw that, seems like a GitHub bug... On the main page, 4.1 is the latest release, when I go to releases, 4.0 is the latest release.

Indeed, this is quite strange behavior on GiHub's end!

[wtfsck]

Should work now. Maybe the bug was caused by the fact that I first selected the v4.0.0 tag and then changed it to v4.1.0.