[LLVMbugs] [Bug 8319] New: clang -O can't build ruby trunk

[bugzilla-daemon at llvm.org]

http://llvm.org/bugs/show_bug.cgi?id=8319

           Summary: clang -O can't build ruby trunk
           Product: clang
           Version: trunk
          Platform: PC
        OS/Version: FreeBSD
            Status: NEW
          Severity: normal
          Priority: P
         Component: -New Bugs
        AssignedTo: unassignedclangbugs at nondot.org
        ReportedBy: naruse at airemix.jp
                CC: llvmbugs at cs.uiuc.edu


clang -O can't build ruby trunk, but clang -O0 can build ruby trunk.
So it seems some optimization breaks ruby.

Ruby's code can be obtain from http://svn.ruby-lang.org/
and I tried with following:

--
Ruby: r29419 on x86_64-freebsd8.1

clang version 2.9 (trunk 115816)
Target: x86_64-unknown-freebsd8.1
Thread model: posix
--

And its build process will crashes like following:

--
snipped...
/home/naruse/local/llvm/bin/clang -O -g -Wall -pedantic -Wno-long-long -Wno-gnu
-Wno-unknown-pragmas  -fPIC -I. -I.ext/include/x86_64-freebsd8.1
-I../../src/ruby-trunk/include -I../../src/ruby-trunk -include ruby/config.h
-include ruby/missing.h -fvisibility=hidden -DRUBY_EXPORT   -o dmyext.o -c
../../src/ruby-trunk/dmyext.c
/home/naruse/local/llvm/bin/clang -O -g -Wall -pedantic -Wno-long-long -Wno-gnu
-Wno-unknown-pragmas  -fPIC -L.  -rdynamic   main.o dmydln.o dmyencoding.o
dmyversion.o miniprelude.o array.o  bignum.o  class.o  compar.o  complex.o 
dir.o  dln_find.o  enum.o  enumerator.o  error.o  eval.o  load.o  proc.o 
file.o  gc.o  hash.o  inits.o  io.o  marshal.o  math.o  node.o  numeric.o 
object.o  pack.o  parse.o  process.o  random.o  range.o  rational.o  re.o 
regcomp.o  regenc.o  regerror.o  regexec.o  regparse.o  regsyntax.o  ruby.o 
safe.o  signal.o  sprintf.o  st.o  strftime.o  string.o  struct.o  time.o 
transcode.o  util.o  variable.o  compile.o  debug.o  iseq.o  vm.o  vm_dump.o 
thread.o  cont.o  ascii.o us_ascii.o unicode.o utf_8.o  newline.o  close.o
dmyext.o -lthr -lrt -lcrypt -lm   -o miniruby
clang: warning: argument unused during compilation: '-g'
clang: warning: argument unused during compilation: '-rdynamic'
/usr/local/bin/ld: error in /usr/lib/crtend.o(.eh_frame); no .eh_frame_hdr
table will be created.
<internal:prelude>:1: [BUG] Segmentation fault
ruby 1.9.3dev (2010-10-07 trunk 29420) [x86_64-freebsd8.1]

-- control frame ----------
c:0003 p:0002 s:0006 b:0006 l:000005 d:000005 TOP    <internal:prelude>:1
c:0002 p:---- s:0004 b:0004 l:000003 d:000003 FINISH
c:0001 p:0000 s:0002 b:0002 l:0008e8 d:0008e8 TOP   
---------------------------
-- Ruby level backtrace information ----------------------------------------
<internal:prelude>:1:in `<compiled>'

[NOTE]
You may have encountered a bug in the Ruby interpreter or extension libraries.
Bug reports are welcome.
For details: http://www.ruby-lang.org/bugreport.html

*** Signal 6

Stop in /home/naruse/obj/ruby-clang.
--

The crashed binary, miniruby, is a ruby subset binary to make full ruby's
files.
In my understanding, it is happend on insns.def:868.

--
 860 DEFINE_INSN
 861 trace
 862 (rb_num_t nf)
 863 ()
 864 ()
 865 {
 866     rb_event_flag_t flag = (rb_event_flag_t)nf;
 867 
 868     EXEC_EVENT_HOOK(th, flag, GET_SELF(), 0, 0 /* TODO: id, klass */);
 869 }
--

And this GET_SELF() is:vm_insnhelper.h:146:
#define GET_SELF() (USAGE_ANALYSIS_REGISTER_HELPER(5, 0, GET_CFP()->self))

GET_CFP() is:
112 #define GET_CFP()  (USAGE_ANALYSIS_REGISTER_HELPER(2, 0, REG_CFP))

REG_CFP is:
81 #define REG_CFP (reg_cfp)

This reg_cfp's value is vanished on clang -O, and crash.
If I change Ruby's code as it uses reg_cfp before this, clang seems not
optimize reg_cfp and ruby works well.

--- insns.def    (revision 29420)
+++ insns.def    (working copy)
@@ -864,6 +864,7 @@
 ()
 {
     rb_event_flag_t flag = (rb_event_flag_t)nf;
+    if (!reg_cfp) sleep(1);

     EXEC_EVENT_HOOK(th, flag, GET_SELF(), 0, 0 /* TODO: id, klass */);
 }

-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.