[Bug tdep/31087] New: [gdb/tdep, arm] FAIL: gdb.base/func-ptrs.exp: continue to decr

[Bug tdep/31087] New: [gdb/tdep, arm] FAIL: gdb.base/func-ptrs.exp: continue to decr

[vries at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=31087

            Bug ID: 31087
           Summary: [gdb/tdep, arm] FAIL: gdb.base/func-ptrs.exp: continue
                    to decr
           Product: gdb
           Version: HEAD
            Status: NEW
          Severity: normal
          Priority: P2
         Component: tdep
          Assignee: unassigned at sourceware dot org
          Reporter: vries at gcc dot gnu.org
  Target Milestone: ---

On a pinebook (with gcc defaulting to generating thumb mode insns) I run into:
...
(gdb) PASS: gdb.base/func-ptrs.exp: go back to main from incr, first time
set calc = decr^M
(gdb) PASS: gdb.base/func-ptrs.exp: set calc to decr
continue^M
Continuing.^M
^M
Program received signal SIGBUS, Bus error.^M
decr (i=-1) at /home/rock/gdb/src/gdb/testsuite/gdb.base/func-ptrs.c:33^M
33      {^M
(gdb) FAIL: gdb.base/func-ptrs.exp: continue to decr
...

The calc variable initially is set to point to incr:
...
(gdb) p calc
$2 = (int (*)(int)) 0xaaaaa4f7 <incr>
...
where incr is:
...
000004f6 <incr>:
 4f6:   b580            push    {r7, lr}
 4f8:   b082            sub     sp, #8
 4fa:   af00            add     r7, sp, #0
 4fc:   6078            str     r0, [r7, #4]
 4fe:   f7ff fff3       bl      4e8 <sentinel>
 502:   687b            ldr     r3, [r7, #4]
 504:   3301            adds    r3, #1
 506:   4618            mov     r0, r3
 508:   3708            adds    r7, #8
 50a:   46bd            mov     sp, r7
 50c:   bd80            pop     {r7, pc}
...

So the first instruction is at 0x4f6, but the lowest bit set in the function
pointer indicates we're jumping to a thumb function.

This is not taken into account when doing "set calc = decr".

In fact, we have the same problem if we do "set calc = incr":
...
(gdb) p calc
$1 = (int (*)(int)) 0xaaaaa4f7 <incr>
(gdb) set calc = incr
(gdb) p calc
$2 = (int (*)(int)) 0xaaaaa4f6 <incr>
...

So I guess this is right:
...
(gdb) p incr
$3 = {int (int)} 0xaaaaa4f6 <incr>
...
but this is wrong:
...
(gdb) p &incr
$4 = (int (*)(int)) 0xaaaaa4f6 <incr>
...

Note btw that manually patching in the thumb bit works fine, and likewise for
decr:
...
(gdb) set calc = incr + 1
(gdb) continue
...

-- 
You are receiving this mail because:
You are on the CC list for the bug.