Re: [PATCH v6 0/5] Reroll patches against v1.8.3.1

Alexey Shumkin <Alex.Crezoff@gmail.com> writes:

Thanks for a reroll, but why this change?

The reason I asked you to avoid 8859-5 is because our test do not use that encoding and I do not want to add new dependency to people when they run test. cp1251 shares exactly the same issue, doesn't it? So in that sense, this change does not make anything better.

That is why I asked you if the breakage you are trying to demonstrate about non-ASCII non-UTF8 encoding was specific to Cyrillic/Russian. I do not recall seeing your answer, but what is the right thing to do depend on it.

 - If the answer is yes, then we would need to add dependency either
   way, and 8859-5 is just as fine as cp1251.

 - If the breakage is not specific to Cyrillic, it should be
   reproducible using 8859-1 (latin-1), and our tests already depend
   on 8859-1, so we wouldn't be adding new dependencies on people.