This is the mail archive of the gcc-bugs@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

[Bug rtl-optimization/64003] New: valgrind complains about get_attr_length_nobnd in insn-attrtab.c from i386.md


https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64003

            Bug ID: 64003
           Summary: valgrind complains about get_attr_length_nobnd in
                    insn-attrtab.c from i386.md
           Product: gcc
           Version: 5.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: rtl-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: dmalcolm at gcc dot gnu.org
              Host: x86_64-unknown-linux-gnu
            Target: x86_64-unknown-linux-gnu
             Build: x86_64-unknown-linux-gnu

Created attachment 34057
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=34057&action=edit
Reproducer

When I run the jit testsuite under valgrind I consistently see reports from
here:

Conditional jump or move depends on uninitialised value(s)
   at 0x5721AED: get_attr_length_nobnd(rtx_insn*) (i386.md:6236)
   by 0x5715F3A: insn_min_length(rtx_insn*) (i386.md:510)
   by 0x4F4BE0A: shorten_branches(rtx_insn*) (final.c:1208)
   by 0x4F52A5A: rest_of_handle_shorten_branches() (final.c:4567)
   by 0x4F52AAE: (anonymous
namespace)::pass_shorten_branches::execute(function*) (final.c:4596)
   by 0x522354D: execute_one_pass(opt_pass*) (passes.c:2306)
   by 0x52237C4: execute_pass_list_1(opt_pass*) (passes.c:2358)
   by 0x52237F5: execute_pass_list_1(opt_pass*) (passes.c:2359)
   by 0x52237F5: execute_pass_list_1(opt_pass*) (passes.c:2359)
   by 0x5223832: execute_pass_list(function*, opt_pass*) (passes.c:2369)
   by 0x4E4884F: cgraph_node::expand() (cgraphunit.c:1773)
   by 0x4E48EE9: expand_all_functions() (cgraphunit.c:1909)

I can reproduce this with cc1 with the attached file at -O2 and above:
  valgrind ./cc1 get-attr-length-i386.c -O2

This is with r217427 on x86_64-unknown-linux-gnu, configuring with:
 --enable-valgrind-annotations 


Turning off the writing of #line directives in read-md.c shows that it's at
line 18500 of the generated insn-attrtab.c:

==5819==    at 0xD952E2: get_attr_length_nobnd(rtx_insn*)
(insn-attrtab.c:18500)

somewhere within this monster conditional:

 18493  int
 18494  get_attr_length_nobnd (rtx_insn *insn ATTRIBUTE_UNUSED)
 18495  {
 18496    switch (recog_memoized (insn))
 18497      {
 18498      case 610:  /* *jcc_1 */
 18499        extract_insn_cached (insn);
>18500        if ((((INSN_ADDRESSES_SET_P () ? INSN_ADDRESSES (INSN_UID (GET_CODE (operands[0]) == LABEL_REF ? XEXP (operands[0], 0) : operands[0])) : 0) - (insn_current_reference_address (insn))) >= (-126)) && (((INSN_ADDRESSES_SET_P () ? INSN_ADDRESSES (INSN_UID (GET_CODE (operands[0]) == LABEL_REF ? XEXP (operands[0], 0) : operands[0])) : 0) - (insn_current_reference_address (insn))) < (128)))


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]