NSJSONSerialization returns nil for certain strings with Unicode sequences about libs-base HOT 2 CLOSED

Thanks ... this was an interesting one. The GNUstep unicode implementation is quite strict and the NSJSONSerialization code was creating bad strings because it uses a 64 character buffer to create small strings, appending them to build longer ones, without checking the integrity of the parts it was creating, so when the buffer boundary was in the middle of a surrogate pair this would cause the creation of the strings to fail.
I changed the code to check for this case, and defer the first half of the surrogate pair to the next chunk, so as long as the source JSON doesn't contain invalid escape sequences it should be OK.
Please close this issue if you think the fix is correct.

from libs-base.

Thank you for the quick fix, works great!

from libs-base.