Keep AutoString::length in sync w/ AutoString::data

[Krovatkin]

This fix makes sure that AutoString::length and AutoString::data are in sync.

[Krovatkin]

AutoString::GetLength is used only in Debugger.cpp (namely, https://github.com/Microsoft/ChakraCore/blob/master/bin/ch/Debugger.cpp#L497). If ChakraFull doesn't have any additional dependencies we could do away with AutoString::GetLength

[Krovatkin]

@Cellule what do you think?

[obastemur]

This supposed to cache string length so we don't measure it again and again.. Well, this is not perf critical part of the solution.

LGTM. Yet, could we do it while caching the length ?

[Krovatkin]

@obastemur I pondered a bit on the caching aspect. Unfortunately, this design created enough complexity to produce a bug. Moreover, some uses in WScriptJsrt.cpp still call strlen. All the uses of AutoString seem to be contained in WScriptJsrt.cpp and Debugger.cpp and neither one seems to be perf critical. Lastly, users of AutoString should be able to cache the result of GetLength in most cases in case we ever hit a perf issue.

Anyways, I'll do whatever you guys tell me to.

fyi @arunetm

[Cellule]

@Krovatkin First of all it won't affect ChakraFull since this file is in ch.exe only

I found the source of the bug.
in commit 356656c (aka #3433). the length used to be cached from the ChakraRTInterface::JsCopyString call in AutoString::Initialize. Instead now it is using a local variable instead thus breaking the initial contract of AutoString.

I think we have to use AutoString::length instead of writtenLength in AutoString::Initialize (both CopyString calls) and that should be the extent of the fix.

Thoughts ?

[Krovatkin]

@Cellule, done

[Krovatkin]

@dotnet-bot test Ubuntu static_ubuntu_linux_release please

[Cellule]

Thanks @Krovatkin, I will merge it for you

[Krovatkin]

@Cellule thank you for your help!