#pentagram@irc.freenode.net logs for 17 Jun 2013 (GMT)

Archive Today Yesterday Tomorrow
Pentagram homepage


[00:44:37] <-- nutron has left IRC (Quit: I must go eat my cheese!)
[12:24:30] <-- Lightkey has left IRC (Excess Flood)
[12:25:10] --> Lightkey has joined #pentagram
[14:24:29] <Dominus> hmm, newest Xcode only has clang now as a compiler. no more llvm-gcc (lock stock gcc has vanished for some time)
[14:24:41] <Dominus> so maybe it's time to make Pentagram clang compatible :)
[14:24:50] <wjp> sure
[14:25:02] <Dominus> :)
[14:25:03] <wjp> let me know if you run into anything if you want to take a shot
[14:25:17] <Dominus> it didn't work at all last time I tried
[14:25:31] <Dominus> I think some instructions we pass not being for clang
[14:26:05] <wjp> instructions = command line options?
[14:27:21] <Dominus> some makefile options
[14:27:38] <Dominus> it starts with "error: unknown argument: '-MMD'"
[14:27:43] <wjp> right
[14:28:07] <wjp> shouldn't be too hard to figure out, but would have to figure out how to do dependency tracking with clang
[14:28:24] <Dominus> perhaps all the -Mxx are not compatible
[14:28:47] <Dominus> I see -MQ and -MP as well
[14:29:03] <wjp> they act together to do dependency tracking
[14:29:17] <wjp> so just need to figure out the best way to do that with clang
[14:38:39] <Dominus> so what we do in pentagram makefile.in is automatic vs. big manual dependency adding in exult's?
[14:48:58] <Dominus> btw. for me it's not a big problem right now since I keep the old command line tools around with gcc but a new install of OSX with the upcoming Xcode stuff will not have those
[14:49:26] <wjp> exult is too different
[15:04:22] <Dominus> hmm,, I wonder why llvm's implementation of -MMx didn't make it into clang...
[15:06:28] <Dominus> searches point to --disable-dependency-tracking when using clang :)
[15:07:26] <-- Colourless has left IRC (Ping timeout: 252 seconds)
[19:28:42] --> nutron has joined #pentagram
[20:11:38] <Dominus> wjp: maybe this would be the course of action. To add a "--disable-dependency-tracking" to configure and move the -Mx stuff to one rule and switch it with that configure rule
[20:12:56] <Dominus> when I manually take out the "-MMD,"$(*D)/$(DEPDIR)/$(*F).d",-MQ,"$@",-MP" in the makefile.in (three instances), Pentagram compiles and runs with clang
[20:34:53] <wjp> nah
[20:35:00] <wjp> should just detect clang and do it properly
[20:35:44] <Dominus> I tried to find out how to do it properly but I lack the certain knowledge to search the right keywords :)
[20:42:51] <wjp> what do you get with 'c++ --version' ?
[20:44:12] <Dominus> Apple LLVM version 4.2 (clang-425.0.28) (based on LLVM 3.2svn)
[20:46:20] <wjp> hm, there's a bug report for -MMD in clang that says newer versions support it
[20:46:27] <wjp> which of course doesn't help you
[20:46:59] <Dominus> does it say which version?
[20:47:11] <wjp> http://llvm.org/bugs/show_bug.cgi?id=9611
[20:53:20] <Dominus> hmm, let me try whether the actual xcode 5 clang does have it
[20:53:35] <Dominus> which is "Apple LLVM version 5.0 (clang-500.1.58) (based on LLVM 3.3svn)"
[20:55:28] <Dominus> yeah!
[20:55:45] <Dominus> sorry, wjp, seems it turned out to be a non-issue
[21:07:32] <wjp> great :-)
[21:08:41] <Dominus> I was so focused on that it doesn't work that I didn't check with the actual "deal breaker" I said it was :)
[21:14:02] <Dominus> on the plus side, clang gives in depth warnings again
[21:14:22] <Dominus> wjp, if you want to see the arch i386 warnings of clang: http://pastie.org/8053308