On 2 April 2014 14:19, Sven Van Caekenberghe <sven@stfx.eu> wrote:
Hi Igor,
Great catch ! How did you stumble upon this one ?
Just showing Max (my new room mate) to implement utf-8 encoding using Virtual CPU DSL, and using your code as an example.
And ya, my sharp eye with perfect precision immediately detected the problem.... (hope it sounds arrogant enough) ;) Frankly, i don't know. What i think is, that i seen comparison with powers of two numbers in my life so many times, that i trained to automatically see if it correct or not. :)
Anyway, it shows that unless you really test all boundaries explicitly, silly one-off errors like that manage to survive for quite some time. I added a unit test doing just that.
Fixed upstream:
=== Name: Zinc-Character-Encoding-Core-SvenVanCaekenberghe.31 Author: SvenVanCaekenberghe Time: 2 April 2014, 2:14:36.890651 pm UUID: 3f6ae285-2973-4451-833b-16afbd0658d3 Ancestors: Zinc-Character-Encoding-Core-SvenVanCaekenberghe.30
Fixed a one-off error in ZnUTF8Encoder>>#nextPut:toStream (thanks Igor Stasenko) === Name: Zinc-Character-Encoding-Tests-SvenVanCaekenberghe.18 Author: SvenVanCaekenberghe Time: 2 April 2014, 2:15:51.938152 pm UUID: c7048056-52ae-484f-a466-7d537f66690a Ancestors: Zinc-Character-Encoding-Tests-SvenVanCaekenberghe.17
Fixed a one-off error in ZnUTF8Encoder>>#nextPut:toStream (thanks Igor Stasenko)
Added ZnCharacterEncoderTests>>#testUTF8Boundaries ===
Thanks again,
Sven
On 02 Apr 2014, at 13:18, Igor Stasenko <siguctua@gmail.com> wrote:
nextPut: character toStream: stream ... code < 65535 ifTrue: [ ...
should be
code < 65536 ifTrue: [
because 3-byte encoding can fit 16 bits, which is range 0...65535 but not 0...65534
-- Best regards, Igor Stasenko.
-- Best regards, Igor Stasenko.