[00:47:12] --- Dan has become available
[01:13:37] --- kula has left
[02:53:04] --- Simon Wilkinson has left
[04:14:00] --- Simon Wilkinson has become available
[06:01:20] --- mvitale has become available
[06:25:59] --- Simon Wilkinson has left
[06:42:59] --- meffie has become available
[06:54:48] --- meffie has left
[07:56:49] --- kula has become available
[08:18:48] --- squinney has become available
[08:27:56] --- deason has become available
[08:31:08] --- Brandon Allbery has left
[08:31:19] --- Brandon Allbery has become available
[08:36:51] --- Dan has left
[08:43:03] --- Simon Wilkinson has become available
[08:51:51] --- squinney has left
[09:58:21] --- meffie has become available
[10:07:00] --- meffie has left
[10:57:05] <kaduk@mit.edu/barnowl> Do we not have a buildbot that uses separate build and source trees?
[11:02:17] <shadow@gmail.com/barnowl0E4B2EC9> objdir build? probably not. broken again huh?
[11:02:39] <kaduk@mit.edu/barnowl> In more than one place.
[11:04:06] <kaduk@mit.edu/barnowl> I'm still not sure what the right fix is for tubik, though. (It
builds utst_{client,server}.o which need a utst_int.h; I'm not sure
if they should be getting one from ., in which case they need an
explicit dependency, or from TOP_INCDIR, which would be a different
dependency.
[11:05:42] <kaduk@mit.edu/barnowl> I suppose I don't actually need to be shooting myself in the foot with
N^2 runtime until I think I have found all of the broken places,
though.
[11:20:45] <Simon Wilkinson> utst is just test stuff - it should be built locally (if at all)
[11:23:00] --- kula has left
[11:23:17] <kaduk@mit.edu/barnowl> I guess I am suffering from insufficiently-stringent regexps combined
with similar-looking prefixes (ubik/utst).
[11:23:40] <Simon Wilkinson> What are you trying to do with regexps?
[11:24:01] <kaduk@mit.edu/barnowl> search for utst_int.h in the Makefile.in
[11:24:12] <kaduk@mit.edu/barnowl> But I was lazy and only searched for _int.h
[11:24:35] <kaduk@mit.edu/barnowl> So I found the code to install ubik_int.h in TOP_INCDIR and confused
myself.
[11:25:54] <Simon Wilkinson> tubik really should go away entirely. It only exists to provide pthreaded versions of udebug (installed) and utst_server and utst_client (not installed)
[11:32:08] --- Brandon Allbery has left
[12:08:41] <kaduk@mit.edu/barnowl> 8891 has first crack at build fixes; I expect that something is not as
good as it could be, somewhere in there.
[12:48:34] --- meffie has become available
[12:57:47] --- Brandon Allbery has become available
[13:13:23] --- kula has become available
[13:41:54] --- kula has left
[13:42:28] --- kula has become available
[13:44:58] <kaduk@mit.edu/barnowl> Hmm, actually, I wonder if none of the libtool stuff is compatible
with separate objdir builds, given I had to symlink all of those .sym
files...
[13:46:28] <Simon Wilkinson> Hmm. Now that you mention it, that's entirely possible. The la.sym file is directly referenced in the build rule, rather than being one of the prerequisites, so make won't do its magic with it
[13:46:38] <kaduk@mit.edu/barnowl> Yeah, I think so.
[13:48:00] <kaduk@mit.edu/barnowl> We could make it depend on ${srcdir}/foo.la.sym, but how will the
libtool invocation would cope with that...
[13:48:56] <Simon Wilkinson> Or just have the libtool rule do —export-symbols $(srcdir)/$@.sym
[13:49:12] <kaduk@mit.edu/barnowl> Well, both places.
[13:49:20] <kaduk@mit.edu/barnowl> I suppose I can take it for a spin.
[13:49:49] <Simon Wilkinson> I would have thought that —export-symbols will work. Make variable evaluation is lazy, so it doesn't matter if that's defined before or after srcdir gets set up.
[13:50:53] <kaduk@mit.edu/barnowl> Oh, I guess make will search ${srcdir} for doing dependency checking.
[13:52:04] <Simon Wilkinson> It should do, yes. And it will fix the depedency list by appending $(srcdir) if it finds it there. The problem is that we aren't pulling the sym file out of the dependency list, but referencing it directly
[13:52:47] <kaduk@mit.edu/barnowl> Yup, all is clear.  (Unless my test build fails.)
[13:59:54] <kaduk@mit.edu/barnowl> Yup, just the libtool change works for the build.
Now testing a tweak for Marc's concerns, then I'll push the update to
gerrit.
[14:35:15] --- Ken Dreyer has left
[14:37:11] --- Ken Dreyer has become available
[14:57:06] --- mvitale has left
[15:52:23] --- meffie has left
[16:00:51] --- shadow@gmail.com/barnowl0E4B2EC9 has left
[16:01:27] --- shadow@gmail.com/barnowl0E4B2EC9 has become available
[16:19:11] --- Brandon Allbery has left
[16:52:15] --- Brandon Allbery has become available
[17:25:07] --- deason has left
[18:05:12] --- shadow@gmail.com/barnowl0E4B2EC9 has left
[18:20:38] --- mdionne has become available
[19:19:55] --- mdionne has left
[20:33:49] --- shadow@gmail.com/barnowl0E4B2EC9 has become available