[Bug target/84310] New: -falign-{labels,loops,jumps} with value >= 32768+1 cause a segfault

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84310

            Bug ID: 84310
           Summary: -falign-{labels,loops,jumps} with value >= 32768+1
                    cause a segfault
           Product: gcc
           Version: unknown
            Status: UNCONFIRMED
          Keywords: ice-on-valid-code
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: marxin at gcc dot gnu.org
  Target Milestone: ---

All releases I have ICE:

$ cat loop2.c
void
a ()
{  
  for (;;)
    ;
}

void
b ()
{  
  for (;;)
    ;
}

$ valgrind --leak-check=yes --trace-children=yes ./xgcc -B. -O2 -S
-o/dev/stdout  loop2.c -falign-loops=32769

==9402== Conditional jump or move depends on uninitialised value(s)
==9402==    at 0xB87FB5: align_fuzz(rtx_def*, rtx_def*, int, unsigned int)
(final.c:599)
==9402==    by 0xB881C2: insn_current_reference_address(rtx_insn*)
(final.c:661)
==9402==    by 0x16EEAF0: insn_current_length(rtx_insn*) (i386.md:453)
==9402==    by 0xB89C86: shorten_branches(rtx_insn*) (final.c:1448)
==9402==    by 0xB90878: rest_of_handle_shorten_branches() (final.c:4728)
==9402==    by 0xB908C8: (anonymous
namespace)::pass_shorten_branches::execute(function*) (final.c:4757)
==9402==    by 0xEBA842: execute_one_pass(opt_pass*) (passes.c:2497)
==9402==    by 0xEBAB93: execute_pass_list_1(opt_pass*) (passes.c:2586)
==9402==    by 0xEBABC4: execute_pass_list_1(opt_pass*) (passes.c:2587)
==9402==    by 0xEBABC4: execute_pass_list_1(opt_pass*) (passes.c:2587)
==9402==    by 0xEBAC1C: execute_pass_list(function*, opt_pass*)
(passes.c:2597)
==9402==    by 0xA3FC5D: cgraph_node::expand() (cgraphunit.c:2139)
==9402== 
==9402== Use of uninitialised value of size 8
==9402==    at 0xA1AF52: INSN_UID(rtx_def*) (rtl.h:1423)
==9402==    by 0xB87FC6: align_fuzz(rtx_def*, rtx_def*, int, unsigned int)
(final.c:603)
==9402==    by 0xB881C2: insn_current_reference_address(rtx_insn*)
(final.c:661)
==9402==    by 0x16EEAF0: insn_current_length(rtx_insn*) (i386.md:453)
==9402==    by 0xB89C86: shorten_branches(rtx_insn*) (final.c:1448)
==9402==    by 0xB90878: rest_of_handle_shorten_branches() (final.c:4728)
==9402==    by 0xB908C8: (anonymous
namespace)::pass_shorten_branches::execute(function*) (final.c:4757)
==9402==    by 0xEBA842: execute_one_pass(opt_pass*) (passes.c:2497)
==9402==    by 0xEBAB93: execute_pass_list_1(opt_pass*) (passes.c:2586)
==9402==    by 0xEBABC4: execute_pass_list_1(opt_pass*) (passes.c:2587)
==9402==    by 0xEBABC4: execute_pass_list_1(opt_pass*) (passes.c:2587)
==9402==    by 0xEBAC1C: execute_pass_list(function*, opt_pass*)
(passes.c:2597)
...

Problem is that we have in shorten_branches set macro:
#define MAX_CODE_ALIGN 16

which is then used for an array.

Anyway similar to -malign-*:

./xgcc -B. -malign-functions=20 loop2.c
cc1: warning: -malign-functions is obsolete, use -falign-functions
cc1: error: -malign-functions=20 is not between 0 and 16

we should also limit the -falign-* options. Let me do it.