Originally, I added the current section to comply with
the .NET team request of further validating my proposal with the CoreFX tests; although this option was quickly proven inadequate.
The CoreFX tests don't refer to the code in the CoreCLR repository; that's why I had to look for some CoreFX methods similar enough to the ones being modified here (i.e.,
MatchChars
and
ParseNumber
in Number.cs). In fact, I found the exact same methods in the file
FormatProvider.Number.cs (note that
these versions were also modified).
Unfortunately, I could only find one test accounting for the aforementioned code:
the parse test for BigInteger
in System.Runtime.Numerics; this wasn't precisely good news because the proposed improvements are much more noticeable in decimal types. This test was quickly deemed irrelevant. In any case, note that I firstly misinterpreted the not-saying-much results as a consequence of relying on
CoreRun.exe (the Core-based alternative uses this program internally). Bear in mind that all the references in this project to
CoreRun.exe come from equivalent not-necessarily-applicable-anymore ideas; also that
the updated version of Project 8 deals with this specific issue.
By following the advice of a .NET team member, I developed
a much simpler testing program which, despite performing notably worse than the other tests on my computer (but reaching the never-seen-before 10% on his computer!), allowed this PR to be finally accepted and merged.