[Bug fortran/33250] New: bus error compiling dqelg.f in scipy on intel mac

[Bug fortran/33250] New: bus error compiling dqelg.f in scipy on intel mac

[claumann at princeton dot edu]

During gfortran compilation of the integration subpack of scipy-0.5.2.1,
gfortran fails with an "internal compiler error: Bus error" when trying to
compile dqelg.f.

Command line and error info:

gfortran:f77: Lib/integrate/quadpack/dqelg.f
Lib/integrate/quadpack/dqelg.f: In function 'dqelg':
Lib/integrate/quadpack/dqelg.f:1: internal compiler error: Bus error
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://gcc.gnu.org/bugs.html> for instructions.
Lib/integrate/quadpack/dqelg.f: In function 'dqelg':
Lib/integrate/quadpack/dqelg.f:1: internal compiler error: Bus error
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://gcc.gnu.org/bugs.html> for instructions.
error: Command "/usr/local/bin/gfortran -Wall -ffixed-form
-fno-second-underscore -fPIC -O3 -funroll-loops -c -c
Lib/integrate/quadpack/dqelg.f -o
build/temp.macosx-10.4-i386-2.5/Lib/integrate/quadpack/dqelg.o" failed with
exit status 1
vpn1-client-a35:/usr/local/src/scipy-0.5.2.1 claumann$ indqelggfortran -v
-bash: indqelggfortran: command not found
vpn1-client-a35:/usr/local/src/scipy-0.5.2.1 claumann$ gfortran -v
Using built-in specs.
Target: i386-apple-darwin8.10.1
Configured with: ../gcc-4.3-20070810/configure --enable-threads=posix
--enable-languages=fortran
Thread model: posix
gcc version 4.3.0 20070810 (experimental)

Fortran code (dqelg.f):
      subroutine dqelg(n,epstab,result,abserr,res3la,nres)
c***begin prologue  dqelg
c***refer to  dqagie,dqagoe,dqagpe,dqagse
c***routines called  d1mach
c***revision date  830518   (yymmdd)
c***keywords  epsilon algorithm, convergence acceleration,
c             extrapolation
c***author  piessens,robert,appl. math. & progr. div. - k.u.leuven
c           de doncker,elise,appl. math & progr. div. - k.u.leuven
c***purpose  the routine determines the limit of a given sequence of
c            approximations, by means of the epsilon algorithm of
c            p.wynn. an estimate of the absolute error is also given.
c            the condensed epsilon table is computed. only those
c            elements needed for the computation of the next diagonal
c            are preserved.
c***description
c
c           epsilon algorithm
c           standard fortran subroutine
c           double precision version
c
c           parameters
c              n      - integer
c                       epstab(n) contains the new element in the
c                       first column of the epsilon table.
c
c              epstab - double precision
c                       vector of dimension 52 containing the elements
c                       of the two lower diagonals of the triangular
c                       epsilon table. the elements are numbered
c                       starting at the right-hand corner of the
c                       triangle.
c
c              result - double precision
c                       resulting approximation to the integral
c
c              abserr - double precision
c                       estimate of the absolute error computed from
c                       result and the 3 previous results
c
c              res3la - double precision
c                       vector of dimension 3 containing the last 3
c                       results
c
c              nres   - integer
c                       number of calls to the routine
c                       (should be zero at first call)
c
c***end prologue  dqelg
c
      double precision abserr,dabs,delta1,delta2,delta3,dmax1,d1mach,
     *  epmach,epsinf,epstab,error,err1,err2,err3,e0,e1,e1abs,e2,e3,
     *  oflow,res,result,res3la,ss,tol1,tol2,tol3
      integer i,ib,ib2,ie,indx,k1,k2,k3,limexp,n,newelm,nres,num
      dimension epstab(52),res3la(3)
c
c           list of major variables
c           -----------------------
c
c           e0     - the 4 elements on which the computation of a new
c           e1       element in the epsilon table is based
c           e2
c           e3                 e0
c                        e3    e1    new
c                              e2
c           newelm - number of elements to be computed in the new
c                    diagonal
c           error  - error = abs(e1-e0)+abs(e2-e1)+abs(new-e2)
c           result - the element in the new diagonal with least value
c                    of error
c
c           machine dependent constants
c           ---------------------------
c
c           epmach is the largest relative spacing.
c           oflow is the largest positive magnitude.
c           limexp is the maximum number of elements the epsilon
c           table can contain. if this number is reached, the upper
c           diagonal of the epsilon table is deleted.
c
c***first executable statement  dqelg
      epmach = d1mach(4)
      oflow = d1mach(2)
      nres = nres+1
      abserr = oflow
      result = epstab(n)
      if(n.lt.3) go to 100
      limexp = 50
      epstab(n+2) = epstab(n)
      newelm = (n-1)/2
      epstab(n) = oflow
      num = n
      k1 = n
      do 40 i = 1,newelm
        k2 = k1-1
        k3 = k1-2
        res = epstab(k1+2)
        e0 = epstab(k3)
        e1 = epstab(k2)
        e2 = res
        e1abs = dabs(e1)
        delta2 = e2-e1
        err2 = dabs(delta2)
        tol2 = dmax1(dabs(e2),e1abs)*epmach
        delta3 = e1-e0
        err3 = dabs(delta3)
        tol3 = dmax1(e1abs,dabs(e0))*epmach
        if(err2.gt.tol2.or.err3.gt.tol3) go to 10
c
c           if e0, e1 and e2 are equal to within machine
c           accuracy, convergence is assumed.
c           result = e2
c           abserr = abs(e1-e0)+abs(e2-e1)
c
        result = res
        abserr = err2+err3
c ***jump out of do-loop
        go to 100
   10   e3 = epstab(k1)
        epstab(k1) = e1
        delta1 = e1-e3
        err1 = dabs(delta1)
        tol1 = dmax1(e1abs,dabs(e3))*epmach
c
c           if two elements are very close to each other, omit
c           a part of the table by adjusting the value of n
c
        if(err1.le.tol1.or.err2.le.tol2.or.err3.le.tol3) go to 20
        ss = 0.1d+01/delta1+0.1d+01/delta2-0.1d+01/delta3
        epsinf = dabs(ss*e1)
c
c           test to detect irregular behaviour in the table, and
c           eventually omit a part of the table adjusting the value
c           of n.
c
        if(epsinf.gt.0.1d-03) go to 30
   20   n = i+i-1
c ***jump out of do-loop
        go to 50
c
c           compute a new element and eventually adjust
c           the value of result.
c
   30   res = e1+0.1d+01/ss
        epstab(k1) = res
        k1 = k1-2
        error = err2+dabs(res-e2)+err3
        if(error.gt.abserr) go to 40
        abserr = error
        result = res
   40 continue
c
c           shift the table.
c
   50 if(n.eq.limexp) n = 2*(limexp/2)-1
      ib = 1
      if((num/2)*2.eq.num) ib = 2
      ie = newelm+1
      do 60 i=1,ie
        ib2 = ib+2
        epstab(ib) = epstab(ib2)
        ib = ib2
   60 continue
      if(num.eq.n) go to 80
      indx = num-n+1
      do 70 i = 1,n
        epstab(i)= epstab(indx)
        indx = indx+1
   70 continue
   80 if(nres.ge.4) go to 90
      res3la(nres) = result
      abserr = oflow
      go to 100
c
c           compute error estimate
c
   90 abserr = dabs(result-res3la(3))+dabs(result-res3la(2))
     *  +dabs(result-res3la(1))
      res3la(1) = res3la(2)
      res3la(2) = res3la(3)
      res3la(3) = result
  100 abserr = dmax1(abserr,0.5d+01*epmach*dabs(result))
      return
      end


-- 
           Summary: bus error compiling dqelg.f in scipy on intel mac
           Product: gcc
           Version: 4.3.0
            Status: UNCONFIRMED
          Severity: critical
          Priority: P3
         Component: fortran
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: claumann at princeton dot edu
 GCC build triplet: i386-apple-darwin8.10.1
  GCC host triplet: i386-apple-darwin8.10.1
GCC target triplet: i386-apple-darwin8.10.1


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33250

[Bug fortran/33250] bus error compiling dqelg.f in scipy on intel mac

[claumann at princeton dot edu]

------- Comment #1 from claumann at princeton dot edu  2007-08-30 17:25 -------
Compiles with no errors using gfortran 4.2.1 for i686-apple-darwin8.


-- 

claumann at princeton dot edu changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |claumann at princeton dot
                   |                            |edu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33250

[Bug fortran/33250] bus error compiling dqelg.f in scipy on intel mac

[pinskia at gcc dot gnu dot org]

-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|critical                    |normal


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33250

[Bug fortran/33250] [Regression 4.3] bus error compiling dqelg.f in scipy on intel mac

[burnus at gcc dot gnu dot org]

------- Comment #2 from burnus at gcc dot gnu dot org  2007-08-30 18:12 -------
(gfortran bugs are by definition not critical or major for GCC as a whole;
still the gfortran team tries to fix regressions as soon as possible.)

I cannot reproduce this problem with today's gfortran on
x86_64-unknown-linux-gnu - neither with -m32 nor with -m64.

Does it also crash if you simply run

   gfortran -c dqelg.f

without -fPIC -O3 -funroll-loops ?

If you have valgrind, could you run gfortran's f951 through valgrind. (To find
the arguments of f951, run gfortran as normal, but pass the "-v" argument; in
the output there is a line which contains a pathname to f951; copy that line.)

If you don't, can you run f951 in gdb and post a backtrace?

Not that it will necessarily help, but it might do so: Could you try a newer
version of gfortran?


-- 

burnus at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |burnus at gcc dot gnu dot
                   |                            |org
            Summary|bus error compiling dqelg.f |[Regression 4.3] bus error
                   |in scipy on intel mac       |compiling dqelg.f in scipy
                   |                            |on intel mac


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33250

[Bug fortran/33250] [Regression 4.3] bus error compiling dqelg.f in scipy on intel mac

[claumann at princeton dot edu]

------- Comment #3 from claumann at princeton dot edu  2007-08-30 18:40 -------
Subject: Re:  [Regression 4.3] bus error compiling dqelg.f in scipy on intel
mac

Howdy-

Thanks for looking into it!

It compiles just fine with

gfortran -c dqelg.f

I'm not especially excited about trying to compile a newer version of  
gfortran at the moment. Do OS X compiles go smoothly? I don't believe  
that valgrind is available for OS X.

This gfortran is available as a binary. I actually downgraded teh  
binary to 4.2 and managed to compile the code. I suppose I could've  
just turned off optimizations as well..

I ran f951 in gdb for you with a backtrace. Don't know if it's  
helpful but here's the output:

penguin:~/gcc-bug claumann$ gdb /opt/usr/local/libexec/gcc/i386-apple- 
darwin8.10.1/4.3.0/f951 GNU gdb 6.3.50-20050815 (Apple version  
gdb-573) (Fri Oct 20 15:50:43 GMT 2006)Copyright 2004 Free Software  
Foundation, Inc.GDB is free software, covered by the GNU General  
Public License, and you are
welcome to change it and/or distribute copies of it under certain  
conditions.
Type "show copying" to see the conditions.
There is absolutely no warranty for GDB.  Type "show warranty" for  
details.
This GDB was configured as "i386-apple-darwin"...Reading symbols for  
shared libraries .... done

(gdb) r dqelg.f -ffixed-form -fPIC -quiet -dumpbase dqelg.f -mmacosx- 
version-min=10.4 -mtune=generic -auxbase dqelg -O3 -version - 
fintrinsic-modules-path /opt/usr/local/bin/../lib/gcc/i386-apple- 
darwin8.10.1/4.3.0/finclude -o /var/tmp//cc3A8yuy.sStarting program: / 
opt/usr/local/libexec/gcc/i386-apple-darwin8.10.1/4.3.0/f951 dqelg.f - 
ffixed-form -fPIC -quiet -dumpbase dqelg.f -mmacosx-version-min=10.4 - 
mtune=generic -auxbase dqelg -O3 -version -fintrinsic-modules-path / 
opt/usr/local/bin/../lib/gcc/i386-apple-darwin8.10.1/4.3.0/finclude - 
o /var/tmp//cc3A8yuy.s
Reading symbols for shared libraries .+ done
GNU F95 version 4.3.0 20070810 (experimental) (i386-apple-darwin8.10.1)
         compiled by GNU C version 4.3.0 20070810 (experimental), GMP  
version 4.2.1, MPFR version 2.2.1.
GGC heuristics: --param ggc-min-expand=30 --param ggc-min-heapsize=4096

Program received signal EXC_BAD_ACCESS, Could not access memory.
Reason: KERN_PROTECTION_FAILURE at address: 0x0000000c
build_classic_dist_vector_1 (ddr=0x434c3b20, ddr_a=0x0,  
ddr_b=0x434af898, dist_v=0x7b4d8c, init_b=0x434e9bd0 "A",  
index_carry=0x7) at ../../gcc-4.3-20070810/gcc/tree-data-ref.c:2719
2719    ../../gcc-4.3-20070810/gcc/tree-data-ref.c: No such file or  
directory.
         in ../../gcc-4.3-20070810/gcc/tree-data-ref.c
(gdb) bt
#0  build_classic_dist_vector_1 (ddr=0x434c3b20, ddr_a=0x0,  
ddr_b=0x434af898, dist_v=0x7b4d8c, init_b=0x434e9bd0 "A",  
index_carry=0x7) at ../../gcc-4.3-20070810/gcc/tree-data-ref.c:2719
#1  0xbfffef44 in ?? ()
Previous frame inner to this frame (corrupt stack?)
(gdb)




On Aug 30, 2007, at 12:12 PM, burnus at gcc dot gnu dot org wrote:

>
>
> ------- Comment #2 from burnus at gcc dot gnu dot org  2007-08-30  
> 18:12 -------
> (gfortran bugs are by definition not critical or major for GCC as a  
> whole;
> still the gfortran team tries to fix regressions as soon as possible.)
>
> I cannot reproduce this problem with today's gfortran on
> x86_64-unknown-linux-gnu - neither with -m32 nor with -m64.
>
> Does it also crash if you simply run
>
>    gfortran -c dqelg.f
>
> without -fPIC -O3 -funroll-loops ?
>
> If you have valgrind, could you run gfortran's f951 through  
> valgrind. (To find
> the arguments of f951, run gfortran as normal, but pass the "-v"  
> argument; in
> the output there is a line which contains a pathname to f951; copy  
> that line.)
>
> If you don't, can you run f951 in gdb and post a backtrace?
>
> Not that it will necessarily help, but it might do so: Could you  
> try a newer
> version of gfortran?
>
>
> -- 
>
> burnus at gcc dot gnu dot org changed:
>
>            What    |Removed                     |Added
> ---------------------------------------------------------------------- 
> ------
>                  CC|                            |burnus at gcc dot  
> gnu dot
>                    |                            |org
>             Summary|bus error compiling dqelg.f |[Regression 4.3]  
> bus error
>                    |in scipy on intel mac       |compiling dqelg.f  
> in scipy
>                    |                            |on intel mac
>
>
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33250
>
> ------- You are receiving this mail because: -------
> You are on the CC list for the bug, or are watching someone who is.
> You reported the bug, or are watching the reporter.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33250

Re: [Bug fortran/33250] [Regression 4.3] bus error compiling dqelg.f in scipy on intel mac

[Andrew Pinski]

On 30 Aug 2007 18:40:51 -0000, claumann at princeton dot edu
<gcc-bugzilla@gcc.gnu.org> wrote:
> build_classic_dist_vector_1 (ddr=0x434c3b20, ddr_a=0x0,
> ddr_b=0x434af898, dist_v=0x7b4d8c, init_b=0x434e9bd0 "A",
> index_carry=0x7) at ../../gcc-4.3-20070810/gcc/tree-data-ref.c:2719
> 2719    ../../gcc-4.3-20070810/gcc/tree-data-ref.c: No such file or
> directory.

This was PR 32946/32573 and was fixed by:
2007-08-24  Jakub Jelinek  <jakub@redhat.com>

        PR tree-optimization/32573
        PR middle-end/32946
        * tree-data-ref.c (initialize_data_dependence_relation): Clear
        DDR_SUBSCRIPTS, DDR_DIR_VECTS and DDR_DIST_VECTS at the beginning.
        (finalize_ddr_dependent): Clear DDR_SUBSCRIPTS after freeing it.
        (build_classic_dist_vector): Return false rather than true if
        DDR_ARE_DEPENDENT is non-NULL at the beginning.  Return false
        if either subscript_dependence_tester_1 or build_classic_dist_vector_1
        returned false.  Don't call save_dist_v before calling
        build_classic_dist_vector_1.
        (free_dependence_relation): Don't guard freeing DDR_SUBSCRIPTS
        with NULL DDR_ARE_DEPENDENT.  Also free DDR_DIST_VECTS and/or
        DDR_DIR_VECTS vectors.

So can you please try a newer version?

Thanks,
Andrew Pinski

[Bug fortran/33250] [Regression 4.3] bus error compiling dqelg.f in scipy on intel mac

[pinskia at gmail dot com]

------- Comment #4 from pinskia at gmail dot com  2007-08-30 18:44 -------
Subject: Re:  [Regression 4.3] bus error compiling dqelg.f in scipy on intel
mac

On 30 Aug 2007 18:40:51 -0000, claumann at princeton dot edu
<gcc-bugzilla@gcc.gnu.org> wrote:
> build_classic_dist_vector_1 (ddr=0x434c3b20, ddr_a=0x0,
> ddr_b=0x434af898, dist_v=0x7b4d8c, init_b=0x434e9bd0 "A",
> index_carry=0x7) at ../../gcc-4.3-20070810/gcc/tree-data-ref.c:2719
> 2719    ../../gcc-4.3-20070810/gcc/tree-data-ref.c: No such file or
> directory.

This was PR 32946/32573 and was fixed by:
2007-08-24  Jakub Jelinek  <jakub@redhat.com>

        PR tree-optimization/32573
        PR middle-end/32946
        * tree-data-ref.c (initialize_data_dependence_relation): Clear
        DDR_SUBSCRIPTS, DDR_DIR_VECTS and DDR_DIST_VECTS at the beginning.
        (finalize_ddr_dependent): Clear DDR_SUBSCRIPTS after freeing it.
        (build_classic_dist_vector): Return false rather than true if
        DDR_ARE_DEPENDENT is non-NULL at the beginning.  Return false
        if either subscript_dependence_tester_1 or build_classic_dist_vector_1
        returned false.  Don't call save_dist_v before calling
        build_classic_dist_vector_1.
        (free_dependence_relation): Don't guard freeing DDR_SUBSCRIPTS
        with NULL DDR_ARE_DEPENDENT.  Also free DDR_DIST_VECTS and/or
        DDR_DIR_VECTS vectors.

So can you please try a newer version?

Thanks,
Andrew Pinski


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33250

[Bug fortran/33250] [Regression 4.3] bus error compiling dqelg.f in scipy on intel mac

[claumann at princeton dot edu]

------- Comment #5 from claumann at princeton dot edu  2007-08-30 18:51 -------
Subject: Re:  [Regression 4.3] bus error compiling dqelg.f in scipy on intel
mac

Great! Will do. Sorry to take your time.
Chris

On Aug 30, 2007, at 12:44 PM, pinskia at gmail dot com wrote:

>
>
> ------- Comment #4 from pinskia at gmail dot com  2007-08-30 18:44  
> -------
> Subject: Re:  [Regression 4.3] bus error compiling dqelg.f in scipy  
> on intel
> mac
>
> On 30 Aug 2007 18:40:51 -0000, claumann at princeton dot edu
> <gcc-bugzilla@gcc.gnu.org> wrote:
>> build_classic_dist_vector_1 (ddr=0x434c3b20, ddr_a=0x0,
>> ddr_b=0x434af898, dist_v=0x7b4d8c, init_b=0x434e9bd0 "A",
>> index_carry=0x7) at ../../gcc-4.3-20070810/gcc/tree-data-ref.c:2719
>> 2719    ../../gcc-4.3-20070810/gcc/tree-data-ref.c: No such file or
>> directory.
>
> This was PR 32946/32573 and was fixed by:
> 2007-08-24  Jakub Jelinek  <jakub@redhat.com>
>
>         PR tree-optimization/32573
>         PR middle-end/32946
>         * tree-data-ref.c (initialize_data_dependence_relation): Clear
>         DDR_SUBSCRIPTS, DDR_DIR_VECTS and DDR_DIST_VECTS at the  
> beginning.
>         (finalize_ddr_dependent): Clear DDR_SUBSCRIPTS after  
> freeing it.
>         (build_classic_dist_vector): Return false rather than true if
>         DDR_ARE_DEPENDENT is non-NULL at the beginning.  Return false
>         if either subscript_dependence_tester_1 or  
> build_classic_dist_vector_1
>         returned false.  Don't call save_dist_v before calling
>         build_classic_dist_vector_1.
>         (free_dependence_relation): Don't guard freeing DDR_SUBSCRIPTS
>         with NULL DDR_ARE_DEPENDENT.  Also free DDR_DIST_VECTS and/or
>         DDR_DIR_VECTS vectors.
>
> So can you please try a newer version?
>
> Thanks,
> Andrew Pinski
>
>
> -- 
>
>
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33250
>
> ------- You are receiving this mail because: -------
> You are on the CC list for the bug, or are watching someone who is.
> You reported the bug, or are watching the reporter.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33250

[Bug fortran/33250] [Regression 4.3] bus error compiling dqelg.f in scipy on intel mac

[pinskia at gcc dot gnu dot org]

------- Comment #6 from pinskia at gcc dot gnu dot org  2007-08-31 01:10 -------
Fixed, already.


-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |FIXED
   Target Milestone|---                         |4.3.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33250