Locale fixes from DragonflyBSD

Review Request #198 - Created June 1, 2016 and updated

Lauri Tirkkonen

Richard PALO

out of curiousity, does this address [any of] the issues involving locales under the perl testsuite?

  1. I don't know what those issues are, so no clue.
  2. build perl then 'gmake check'
    see illumos@ 27/07/2015 "perl 5.22 (32 bit , 64bitint) tests failures"

  3. Based on the mailing list thread you mentioned, I thought this unlikely. However, it does affect the '../lib/locale.t' tests, although not in a good way: now we get a segfault before running all the tests. I have C test case for this (it's in strxfrm), and I need to investigate.
Gordon Ross

usr/src/cmd/localedef/wide.c (Diff revision 1)
There's no reason to obfuscate the initializer that way, is there?
  1. I think you're right. DragonFly did it this way; the intent is to zero the entire length of the buffer, but it looks like that will happen with the string literal initializer as well. I'll change it.
Yuri Pankov

Guess this can be closed, I've taken care of #6904 and #6905, #6906 is still open, but I've fix ready.