The problem with ths fix is that the current system reads the number of bytes and advances the iterator only that number of bytes. So if I were to make this change, I would also have to make another change on the base64 input. But then I would have a problem with reading old archives which woudn't have the extra byte. Nevertheless, I would much like to see a correct, portable and more widely usable version of this iterator. So, If you would like to see this do the following: a) open up a track ticket on this. b) post patch for output AND input c) run tests - at least test_bas64 iterator Then I'll look at rolling it in - along with tweaks to make it backward compatible for existing archives. Let no good deed go unpunished. Robert Ramey Vyacheslav E. Andrejev wrote:
Hello,
Code that pads base64 output with '=' looks in basic_text_oprimitive.ipp like following:
std::size_t padding = 2 - count % 3; if(padding > 1) *oi = '='; if(padding > 2) *oi = '=';
It will give the following transformations of a few simple strings:
"012" -> "MDEy=" "0123" -> "MDEyMw" "01234" -> "MDEyMzQ" "012345" -> "MDEyMzQ1="
Which violates the principle of the base 64 that the encoded output length is multiple of 4 bytes. Right output for the samples above should be:
"012" -> "MDEy" "0123" -> "MDEyMw==" "01234" -> "MDEyMzQ=" "012345" -> "MDEyMzQ1"
The problem with the code above is that the condition (padding > 2) is never satisfied. I think that the proper code should look like
std::size_t padding = 2 - count % 3; if (padding < 2) { *oi = '='; if (padding > 0) *oi = '='; }