01:35:51 <raeburn> Hm, mkey migration still has that stash bug, ticket 6617… looks like it's been around since the original integration.
03:10:05 <ghudson> Yes, that hasn't been fixed.
06:59:26 <raeburn> Think I've got that fixed now.
07:01:05 <raeburn> The hard-link idea may not be so good, I just remembered — one might try to build in AFS.  Falling back to cp may be necessary, at least if ln doesn't work.
07:03:06 <raeburn> I've made "variable is used uninitialized" an error for gcc versions that support it, now that 6617 is dealt with.
09:24:44 <raeburn> Ah, I broke the build; didn't get all the test programs with uninitialized warnings.
12:21:04 <ghudson> I thought about build trees crossing devices and figured it wasn't common enough to worry about.  I didn't think about build trees in AFS.
12:22:00 <ghudson> It would probably be simplest to make copies, although that makes the name "liblinks" somewhat deceptive.
13:36:59 <ghudson> I created a wiki page at http://k5wiki.kerberos.org/wiki/Portability_research and added it to the lore compatibility.  It has notes on MSVC's C99 support and also the results of my old platform survey for IPv4/IPv6 pktinfo support.