Le quartidi 4 brumaire, an CCXXIV, wm4 a écrit : > Well, I'm not sure if this only affects LC_NUMERIC and maybe LC_COLLATE?
I grant you that LC_PAPER or LC_TELEPHONE are probably harmless; they were added after I made my mental list. It should work by whitelist, not by blacklist. > On the other hand, setting LC_CTYPE could have some bad effects too. Very true. Unfortunately, some genius decided to use LC_CTYPE to determine the tty's encoding, instead of $TERM, this makes LC_CTYPE unavoidable. > I'd be fine with that, but then we should maybe check the current > locale for sanity at certain points, and raise a clear error if it's > broken. > > Many programs set locale by default, including GUI toolkits like Qt (!). > Qt in particular document this explicitly: > > http://doc.qt.io/qt-5/qcoreapplication.html#locale-settings > > So we shouldn't silently break everything. But again, breaking loudly > is ok IMHO. Agree. > You could demand the same from FFmpeg. It's impractical because not all > systems support these functions. Exactly. Printing number with a localized separator is a minor and optional feature, and rather easy to reimplement in the application. Parsing floating point numbers is used everywhere and hard to do if you want to keep the full precision. Which one should work by default is really a no-brainer. (Well, actually, I consider localizing the decimal separator to be completely stupid. Numbers are numbers.) > I think we don't need to block the patch on this issue anymore, because: Agree too. Regards, -- Nicolas George
signature.asc
Description: Digital signature
_______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel