[Bug target/23224] New: Severe outages with -fPIC

[Bug target/23224] New: Severe outages with -fPIC

[jkj at sco dot com]

As a general part of testing for my platform (UnixWare),
I always run teh testsuites with -fPIC (I also multilib
for -fPIC). There is a general problem with PIC, and
its not unique to UnixWare, many of the same tests fail
on Linux when running the test suites with -fPIC. Many
of the errors are along the lines of:

   cant find a register in class 'GENERAL_REGS' while
   reloading 'asm'

For some of the test cases below, I believe its simply
becuase of the way the test case is written.  However,
some I think are more problematic. They compile OK but
dont run when compiled -fPIC, but run fine when compiled
without it. In the list of test cases below, they are
marked with '*'. Those marked with '+' are the ones
that eject the GENERAL_REGS error above. The ones
marked with 'L' fail on Linux too.

+ L gcc.c-torture/compile/20000804-1.c -O0
+ L gcc.c-torture/compile/20000804-1.c -O1
 *  gcc.c-torture/execute/20050121-1.c -O0
 *  gcc.c-torture/execute/20050121-1.c -O1
 *  gcc.c-torture/execute/20050121-1.c -O2
 *  gcc.c-torture/execute/20050121-1.c -O3 -fomit-frame-pointer
 *  gcc.c-torture/execute/20050121-1.c -O3 -g
 *  gcc.c-torture/execute/20050121-1.c -Os
 *  gcc.c-torture/execute/960512-1.c -O0
 *  gcc.c-torture/execute/960512-1.c -O1
 *  gcc.c-torture/execute/960512-1.c -O2
 *  gcc.c-torture/execute/960512-1.c -O3 -fomit-frame-pointer
 *  gcc.c-torture/execute/960512-1.c -O3 -g
 *  gcc.c-torture/execute/960512-1.c -Os
 *  gcc.c-torture/execute/complex-1.c -O0
 *  gcc.c-torture/execute/complex-1.c -O1
 *  gcc.c-torture/execute/complex-1.c -O2
 *  gcc.c-torture/execute/complex-1.c -O3 -fomit-frame-pointer
 *  gcc.c-torture/execute/complex-1.c -O3 -g
 *  gcc.c-torture/execute/complex-1.c -Os
 *  gcc.c-torture/execute/complex-2.c -O0
 *  gcc.c-torture/execute/complex-2.c -O1
 *  gcc.c-torture/execute/complex-2.c -O2
 *  gcc.c-torture/execute/complex-2.c -O3 -fomit-frame-pointer
 *  gcc.c-torture/execute/complex-2.c -O3 -g
 *  gcc.c-torture/execute/complex-2.c -Os
 *  gcc.c-torture/execute/complex-6.c -O0
 *  gcc.c-torture/execute/complex-6.c -O1
 *  gcc.c-torture/execute/complex-6.c -O2
 *  gcc.c-torture/execute/complex-6.c -O3 -fomit-frame-pointer
 *  gcc.c-torture/execute/complex-6.c -O3 -g
 *  gcc.c-torture/execute/complex-6.c -Os
 *  gcc.dg/compat/scalar-return-3 c_compat_x_tst.o-c_compat_y_tst.o

gcc.dg/compat/vector-1_x.c
gcc.dg/compat/vector-2_x.c
(insn 48 47 34 (set (reg:SI 21 xmm0)
 (mem:SI (plus:SI (reg/f:SI 6 bp)
  (const_int -44 [0xffffffd4)) [0 S4 A8] 37 {*movsi_1_nointerunit} (nil)
(nil)))
ICE: in extract_constrain_insn_cached, at recog.c:2000

gcc.dg/compat/vector-1_y.c compile fails with:
In function test_v16qi:9 unable to find a register to spill in
class `GENERAL_REGS'. This is the insn:
(insn 45 49 46 0 (set (subreg:SI (reg:V16QI 73) 0)
  (subreg:SI (reg:V16QI 21 xmm0) 0)) 37 {*movsi_1_nointerunit} (nil)
 (nill))

gcc.dg/compat/vector-2_y.c compile fails with:
In function test_v4sf:9 unable to find a register to spill in
class `GENERAL_REGS'. This is the insn:
(insn 45 49 46 0 (set (subreg:SI (reg:V4SF 73) 0)
  (subreg:SI (reg:V16QI 21 xmm0) 0)) 37 {*movsi_1_nointerunit} (nil)
 (nill))

gcc.dg/20011119-1.c fails (also on Linux) with:
 20011119-1.c:69: warning: asm operand 0 probably doesn't match constraints
 20011119-1.c:69: error: impossible constraint in asm

gcc.dg/20020919-1.c fails (also on Linux) with:
 error: PIC register '%ebx' clobbered by asm
gcc.dg/clobbers.c fails the same way

gcc.dg/i386-387-3.c
gcc.dg/i386-387-4.c
gcc.dg/i386-cmov1.c
gcc.dg/i386-mul.c
gcc.dg/i386-ssetype-1.c
gcc.dg/i386-ssetype-3.c
gcc.dg/i386-ssetype-5.c
  All fail in assemble scans because the test cases dont take
  PIC into account. This isnt a compiler error, but a test
  case error. All fail on Linux too.

gcc.dg/sibcall-6.c execution test fails (on linux too)

I suspect that -fPIC doesn't get nearly as much testing as
it should, and I also bet that there is a solution to almost
all of these thats going to turn out to be pretty simple,
and will probably solve them all, as they mostly seem very
related.

-- 
           Summary: Severe outages with -fPIC
           Product: gcc
           Version: 3.4.4
            Status: UNCONFIRMED
          Severity: critical
          Priority: P1
         Component: target
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: jkj at sco dot com
                CC: gcc-bugs at gcc dot gnu dot org
 GCC build triplet: i586-pc-sysv5 i686-pc-linux-gnu
  GCC host triplet: i586-pc-sysv5 i686-pc-linux-gnu
GCC target triplet: i586-pc-sysv5 i686-pc-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23224

[Bug target/23224] Severe outages with -fPIC

[pinskia at gcc dot gnu dot org]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
 GCC target triplet|                            |i586-pc-sysv5


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23224

[Bug target/23224] Severe outages with -fPIC

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-08-04 03:41 -------
20000804-1.c is PR 19227.
20011119-1.c is PR 19228.
20020919-1.c is PR 19275
sibcall-6.c is PR 19278
The above testcases are just testcase problems and nothing more.

You have to look into why the other testcase fails because besides the one about don't fail on linux 
unless it is a testcase issue.

A lot of the sse/mmx issues have been fixed in 4.0.0 so you might want to double check there.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
  BugsThisDependsOn|                            |19227, 19228, 19275, 19278
           Severity|critical                    |normal
  GCC build triplet|i586-pc-sysv5 i686-pc-linux-|
                   |gnu                         |
   GCC host triplet|i586-pc-sysv5 i686-pc-linux-|
                   |gnu                         |
 GCC target triplet|i586-pc-sysv5 i686-pc-linux-|
                   |gnu                         |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23224

[Bug target/23224] Severe outages with -fPIC

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-08-04 03:43 -------
Looks like complex returns are broken for your target with -fPIC.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23224

[Bug target/23224] Severe outages with -fPIC

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-08-04 03:48 -------
Some more testcase failures with -fPIC:
[3.4 only] ICE in g++.old-deja/g++.pt/asm1.C and asm2.C with -fpic/-fPIC (PR 19226)
Failure in gcc.dg/assign-warn-3.c on i686-pc-linux-gnu with -fpic/-fPIC (PR 19232)
Execute failure in gcc.c-torture/execute/builtins/strlen-3.c with -fpic/-fPIC (PR 19231)

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
  BugsThisDependsOn|                            |19226, 19231, 19232


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23224

[Bug target/23224] Severe outages with -fPIC

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-08-04 03:50 -------
And -fPIC is tested:
http://gcc.gnu.org/ml/gcc-testresults/2005-08/msg00192.html (3.4 branch)
http://gcc.gnu.org/ml/gcc-testresults/2005-08/msg00168.html (4.0 branch)
http://gcc.gnu.org/ml/gcc-testresults/2005-08/msg00167.html (mainline)

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23224

[Bug target/23224] Severe outages with -fPIC

[jkj at sco dot com]

------- Additional Comments From jkj at sco dot com  2005-08-04 04:03 -------
Wow. You sure are on top of things Andrew :)

Sorry I missed the -fPIC reports. The few I looked at didn't have them.

As for complex returns being broken on my target, do you by any chance have any
insight as to where I may start looking? Even the vaguest of clue will head me
in the right direction. TIA.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23224

[Bug target/23224] Severe outages with -fPIC

[steven at gcc dot gnu dot org]

------- Additional Comments From steven at gcc dot gnu dot org  2005-08-04 23:55 -------
Known. 
 

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|                            |1
   Last reconfirmed|0000-00-00 00:00:00         |2005-08-04 23:55:41
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23224