bugGNU Octave - Bugs: bug #52607, test suite hangs or fails after...

 
 

bug #52607: test suite hangs or fails after recent csets

Submitter:  Rik <rik5>
Submitted:  Wed 06 Dec 2017 06:46:04 PM UTC
   
 
Category:  Interpreter Severity:  3 - Normal
Priority:  5 - Normal Item Group:  Regression
Status:  Fixed Assigned to:  jwe
Originator Name:  Open/Closed:  * Closed
Release:  * dev Operating System:  * GNU/Linux
Fixed Release:  None Planned Release:  None
* Mandatory Fields

Add a New Comment Rich Markup
   

Jump to the original submission

Thu 07 Dec 2017 11:54:08 PM UTC, comment #20: 

Nevermind -- it appears an opeblas problem.

Dmitri.

Dmitri A. Sergatskov <dasergatskov>
Thu 07 Dec 2017 11:49:04 PM UTC, comment #19: 

Now I started getting sporadic hangs during "make check" at
classes.tst (never when run test classes.tst at the prompt)...

I think it had happened before, but I forgot how it was resolved.

Dmitri.
--


Dmitri A. Sergatskov <dasergatskov>
Thu 07 Dec 2017 08:14:09 PM UTC, comment #18: 

And another change for mxe-octave so that the general package should build again:

  http://hg.octave.org/mxe-octave/rev/5e3fb4c40644

John W. Eaton <jwe>
Group administrator
Thu 07 Dec 2017 05:07:58 PM UTC, comment #17: 

I pushed a fix for the OS X build problem:

http://hg.savannah.gnu.org/hgweb/octave/rev/51a1d6c84d0c

John W. Eaton <jwe>
Group administrator
Thu 07 Dec 2017 05:06:24 PM UTC, comment #16: 

clang-3.9-debian is failing due to osmesa

Dmitri.

Dmitri A. Sergatskov <dasergatskov>
Thu 07 Dec 2017 04:49:12 PM UTC, comment #15: 

Also mxe builds are broken because of symbol table changes that cause build failures for {un,}packfields.  Those are horrible functions that deserve to fail.  But I will see about fixing them anyway.

John W. Eaton <jwe>
Group administrator
Thu 07 Dec 2017 04:36:01 PM UTC, comment #14: 

Yay!  This bug seems to be fixed.  I took a look at the buildbots and they are mostly happy now.  The failures are with old versions of clang, clang on Mac OSX, or due to OSmesa issues which we already know about.


Rik <rik5>
Group administrator
Thu 07 Dec 2017 06:39:56 AM UTC, comment #13: 

I pushed the following changeset:

  http://hg.savannah.gnu.org/hgweb/octave/rev/ea3458c1d884

It seems to fix the problem with decic.m tests hanging, at least for me.  I've tried it with the address sanitizer and a regular build with and without the JIT compiler enabled (though now I don't think that anything to do with it).  I hope the buildbots will also be happy with this change.

John W. Eaton <jwe>
Group administrator
Thu 07 Dec 2017 05:17:39 AM UTC, comment #12: 

I think ENABLE_64 is intended.  The NEWS file for 4.4 contains


 ** On systems with 64-bit pointers, --enable-64 is now the default and
    Octave always uses 64-bit indexing.  However, if the configure
    script determines that the BLAS library uses 32-bit integers, then
    operations using the following libraries are limited to arrays with
    dimensions that are smaller than 2^31 elements:

      BLAS  LAPACK  QRUPDATE  SuiteSparse  ARPACK



Rik <rik5>
Group administrator
Thu 07 Dec 2017 03:03:57 AM UTC, comment #11: 

I also see

ENABLE_64 = 1

(which I did not enable) -- I assume it is a similar buglet.

"make check" hangs intermittently on my computer(s).

Dmitri.
--

Dmitri A. Sergatskov <dasergatskov>
Thu 07 Dec 2017 02:42:07 AM UTC, comment #10: 

Woops, you found a bug in the configure script.  It was just checking whether ENABLE_JIT was defined (even though it was defined to "no") before calling AC_DEFINE.  I fixed that on the stable branch here http://hg.savannah.gnu.org/hgweb/octave/rev/80948d6e58c2.

Unfortunately, I still hang in decic.m.

Rik <rik5>
Group administrator
Thu 07 Dec 2017 02:06:38 AM UTC, comment #9: 

That is what I do, but then:

octave:2> _octave_config_info_ ("ENABLE_JIT")
ans = 1

Dmitri A. Sergatskov <dasergatskov>
Thu 07 Dec 2017 02:01:07 AM UTC, comment #8: 

@Dmitri: It isn't the default for me.  I just did 'make distclean' followed by './configure' and JIT was disabled.

Rik <rik5>
Group administrator
Thu 07 Dec 2017 01:27:41 AM UTC, comment #7: 

BTW, I just noticed that JIT is enabled by default now.
When did this happen and was this intentional?

Dmitri.
--

Dmitri A. Sergatskov <dasergatskov>
Wed 06 Dec 2017 09:45:42 PM UTC, comment #6: 

Maybe I was confused before.  This problem seems to be a bit random.  It's now hanging for me with the JIT compiler disabled.


John W. Eaton <jwe>
Group administrator
Wed 06 Dec 2017 09:12:40 PM UTC, comment #5: 

I'm running a default configure as well which means that JIT is disabled.  Interestingly, if I run


test decic


by itself it passes.  The problem is only during the long run of the test suite.

Rik <rik5>
Group administrator
Wed 06 Dec 2017 08:17:36 PM UTC, comment #4: 

After

http://buildbot.octave.org:8010/changes/3981

<<<<
Backed out changeset 4e746afab617

The fix for this problem is not as simple as it would appear...

>>>>


Fedora's buildbot are OK.

Dmitri.

Dmitri A. Sergatskov <dasergatskov>
Wed 06 Dec 2017 08:15:50 PM UTC, comment #3: 

All buildbots are having problems. They are configured with def
options which means JIT off.

Dmitri.

Dmitri A. Sergatskov <dasergatskov>
Wed 06 Dec 2017 07:39:06 PM UTC, comment #2: 

Are you building with the JIT compiler enabled?  The decic tests also hang for me if it is enabled but not if it is disabled.

I know that is not a solution for the problem.

John W. Eaton <jwe>
Group administrator
Wed 06 Dec 2017 07:01:48 PM UTC, comment #1: 

The cset that introduces the hang in decic.m is


changeset:   24361:8bcfddad15ec
user:        John W. Eaton <jwe@octave.org>
date:        Mon Nov 27 01:12:05 2017 -0500
summary:     use shared_ptr to manage symbol_scope objects



Rik <rik5>
Group administrator
Wed 06 Dec 2017 06:46:04 PM UTC, original submission:  

Running 'make check' works with cset 24360


changeset:   24360:ff4717c3223f
parent:      24356:8b14ba8296af
parent:      24359:1da659b26bec
user:        Rik <rik@octave.org>
date:        Tue Dec 05 11:35:11 2017 -0800
summary:     maint: merge of stable to default.


The problem seems to have been introduced by one of these 3 csets:


changeset:   24363:4e746afab617
user:        John W. Eaton <jwe@octave.org>
date:        Wed Dec 06 11:45:44 2017 -0500
summary:     don't store relative directory names in load path (bug #52598)

changeset:   24362:3fc24b792a24
user:        John W. Eaton <jwe@octave.org>
date:        Tue Dec 05 17:09:24 2017 -0500
summary:     avoid including symtab.h, symscope.h, or symrec.h unnecessarily

changeset:   24361:8bcfddad15ec
user:        John W. Eaton <jwe@octave.org>
date:        Mon Nov 27 01:12:05 2017 -0500
summary:     use shared_ptr to manage symbol_scope objects




Rik <rik5>
Group administrator

 

(Note: upload size limit is set to 16384 kB, after insertion of the required escape characters.)

Attach Files:
   
   
Comment:
   

No files currently attached

 

Depends on the following items: None found

Items that depend on this one: None found

 

Carbon-Copy List
  • -email is unavailable- added by dasergatskov (Posted a comment)
  • -email is unavailable- added by jwe (Posted a comment)
  • -email is unavailable- added by rik5 (Submitted the item)
  •  

    There are 0 votes so far. Votes easily highlight which items people would like to see resolved in priority, independently of the priority of the item set by tracker managers.

    Only group members can vote.

     

    Follow 3 latest changes.

    Date Changed by Updated Field Previous Value => Replaced by
    2017-12-07 rik5 StatusConfirmed Fixed
        Open/ClosedOpen Closed
    2017-12-07 rik5 StatusNone Confirmed

    Back to the top

    Powered by Savane 3.13-d3ae.
    Corresponding source code