Regex parser regression

[jacobzed]

Description

The following regex doesn't match in .net 8. The same regex matches in framework 4.72.

Reproduction Steps

var re = new Regex(@"^(.+?) (says?),\s'(.+)'$", RegexOptions.RightToLeft);
var sample = "User says, 'adventure'";
var match = re.Match(sample);
Console.WriteLine(match.Success); // should return true like in 4.72

Expected behavior

Regex match success

Actual behavior

Regex match fails

Regression?

Yes, this works in Framework 4.72

Known Workarounds

Removing RegexOptions.RightToLeft, or putting "s?" outside capturing group:
var re = new Regex(@"^(.+?) (say)s?,\s'(.+)'$", RegexOptions.RightToLeft);

Configuration

No response

Other information

No response

[dotnet-policy-service]

Tagging subscribers to this area: @dotnet/area-system-text-regularexpressions
See info in area-owners.md if you want to be subscribed.

[steveharter]

Did verify the repro. .NET v6 returns true while v7 returns false, so the behavior changed in v7.

The RegexOptions.RightToLeft is causing this to fail.

[steveharter]

cc @stephentoub

[stephentoub]

Looks like this was introduced as part of an optimization that shouldn't be applying to RTL. We're looking to see whether we can combine loops with a multi that comes after it (a multi follows a loop in the example pattern because RTL ends up reversing concatenations), but because of the RTL we're looking at the wrong end of the multi, so it ends up erroneously combining the s? with the s that begins "say".