> On Dec 17, 2014, at 7:41 PM, Marshall Clow <[email protected]> wrote: > > >> On Dec 11, 2014, at 8:22 AM, Ed Schouten <[email protected] >> <mailto:[email protected]>> wrote: >> >> 2014-12-11 15:17 GMT+01:00 Joerg Sonnenberger <[email protected] >> <mailto:[email protected]>>: >>> Sane system use the same logic for the default C locale and the >>> non-default locales, just with a different table. >> >> Sane systems also tend to evolve over time. The snippet of code that >> we use for FreeBSD, for example, will not work on a system from before >> November 2011. Who knows what Microsoft is going to do with MSVCRT.DLL >> in the future? >> >> That said, I am not the maintainer of libc++. Feel free to make any >> modifications to the proposed patch before committing. > > I’ve been discussing this with Dan on IRC. > > I am quite happy to add a default rune table to libc++ (conditionally > included) and to use it on systems that lack (or have defective) tables of > their own. > > I am less enthusiastic about libc++ always providing its’ own - at this time. > I’m worried about compatibility issues.
Finally coming back to this. The first step is to add tests. :-) I’ve added tests for the contents of ctype<char>::classic_table. (commit 231077) I’ll be watching the buildbots. Next step: Add a default table, and test that. Third step: Let people turn on the default table for their systems. Is everyone ok with that? — Marshall
_______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
