Uncaught exception in decoding about gd.py

[nekitdev]

Point is, ignoring unknown bytes gives garbage to the programmers using the library (and consequently to their projects). I think we need to figure out the best way to handle encoding/decoding issues before applying changes.

[QuasiStellar]

gives garbage to the programmers

Isn't it better than not giving anything and crashing with an error instead?

[nekitdev]

I said we need to find the best way — the proposed solution is merely better.

[QuasiStellar]

Well, I wasted multiple hours investigating this crash and then rerunning the code that depends on it. None of this would happen if the error-handling strategy was set to "replace" or "ignore" (which GD itself is using) from the start. The problem affects 1 very old level out of 38 000 rated ones and I really don't care if its description contains � instead of a non-breaking space that was supposed to be there for whatever reason. I do care when the libraries I use break down for such minor reasons though.

[QuasiStellar]

In general I think that methods such as search_levels should never throw if they can get at least some valid info. If something is not right (such as in this case), a warning can be printed.

[nekitdev]

You can now handle this via setting GD_ENCODING_ERRORS environment variable.