Let me approach this differently, and hope I make more sense The values objects and the C standard should largely agree:
1. If we're using c99 or above we want values-xpg6 2. If we're using c89, we only want xpg4 if we're trying to be xpg4 3. If no standard is requested, we want values-Xa, and no values-xpg I think my patch gets #1 wrong where the standard is allowed to default, I thought the spec language accounted for that but now I'm not so sure. I think this is what you meant. My patch ignores #2, it doesn't default to values-xpg4 for c89, but it also doesn't leave you with a way to _ask_ for values-xpg4. GCC 8 defaults values-xpg4 for c89, which has the opposite problem for to my patch, in that it breaks #3 which would cause an amount of chaos for illumos but likely would not have any effect on end users. The problem with #2/#3 is that we generally hold to the traditional behaviour (and have /usr/xpg4, and /usr/xpg6). Not being able to escape the xpg values would leave the /usr/bin versions being xpg4 -- or at least more xpg4 than previously. I've Cc'd Robert and John, in the hopes that both they can figure out what I'm trying to say, and that they also have solutions (or can just tell me I'm wrong). ------------------------------------------ illumos: illumos-discuss Permalink: https://illumos.topicbox.com/groups/discuss/T143f32047bf3c2fe-M4dc9673a6347aec8f1febbe9 Delivery options: https://illumos.topicbox.com/groups/discuss/subscription
