TextBuffer contentsToWriter does not validate string length

[pjfanning]

The StreamReadConstraints maxStringLen may kick in when the buffer size is updated but you can get scenarios where like with contentsAsString, you need to check at the end too. The maxStringLen is documented as approximate, so we could choose not to worry about this extra use case.

Ironically, since contentsToWriter already supports throwing IOException, this method is better suited to reporting exceptions that methods like contentsAsString.

[cowtowncoder]

Ok good to note. It looks like it probably would make sense to validate it here.

I did also realize that we are actually not validating sizes on various resetXxx() calls for shared buffers, but we are waiting for accessor. If validation was done upfront, we could avoid checks for shared buffers later on. That probably won't have much performance impact tho.

However... thinking further, I think it should never be necessary to check lengths on contentAsString() since there must have been one call to "finish" contents before it gets called: one of:

• setCurrentAndReturn()
• finishCurrentSegment() (intermediate buffer fill, not necessarily last)
• finishAndReturn() (relies on contentAsString() for its check)

so I think that checks should ideally be moved into only these 3, plus appends, resetXxx(), and removed from accessors.
There's still need to change signatures I suppose.

EDIT: above is incorrect as there were append() methods for which final check was needed. And trying to avoid check in contentsAsString() ended up not being all that useful in the end, so checks remain in there and other places.

Whether this issue should be tackled is different story; will leave open for possibly addressing in future.