bugPSPP - Bugs: bug #44877, linear regression... program crash

 
 

bug #44877: linear regression... program crash

Submitter:  None
Submitted:  Sun 19 Apr 2015 07:26:36 AM UTC
   
 
Category:  Other Severity:  7 - Major
Status:  Fixed Assigned to:  blp
Open/Closed:  Closed Release:  None
Effort:  0.00
* Mandatory Fields

Add a New Comment Rich Markup
   

Jump to the original submission

Mon 04 Jan 2016 04:14:06 AM UTC, comment #11: 

I fixed this.

Ben Pfaff <blp>
Group administrator
Thu 12 Nov 2015 05:45:29 PM UTC, comment #10: 

Right.  It appears as if the transaction is presenting us with the wrong caseproto geometry.  But I cannot see why.

John Darrington <jmd>
Group administrator
Thu 12 Nov 2015 04:06:50 AM UTC, comment #9: 

valgrind reports the following:


==9265== Invalid free() / delete / delete[] / realloc()
==9265==    at 0x402A3A8: free (vg_replace_malloc.c:473)
==9265==    by 0x418B3F8: value_destroy (value.h:150)
==9265==    by 0x418B3F8: destroy_long_strings (caseproto.c:399)
==9265==    by 0x418C4BE: case_unref__ (case.c:452)
==9265==    by 0x418F0C7: case_unref (case.h:142)
==9265==    by 0x418F0C7: casewindow_memory_destroy (casewindow.c:238)
==9265==    by 0x418F146: casewindow_destroy (casewindow.c:101)
==9265==    by 0x418F607: casereader_window_destroy (casewriter.c:284)
==9265==    by 0x418E4A8: random_reader_destroy (casereader.c:536)
==9265==    by 0x418E5D7: casereader_destroy (casereader.c:94)
==9265==    by 0x409D805: save_trans_free (regression.c:152)
==9265==    by 0x41BA134: trns_chain_destroy (transformations.c:102)
==9265==    by 0x41963C3: proc_cancel_all_transformations (dataset.c:806)
==9265==    by 0x4196690: dataset_clear (dataset.c:225)
==9265==    by 0x41966EA: dataset_destroy (dataset.c:201)
==9265==    by 0x41AC44F: session_destroy (session.c:71)
==9265==    by 0x804A445: main (main.c:164)
==9265==  Address 0x271e629c is not stack'd, malloc'd or (recently) free'd
==9265==


Ben Pfaff <blp>
Group administrator
Fri 23 Oct 2015 04:16:23 PM UTC, comment #8: 

I had another look at this, and here is a simpler test case:
It seems that the cases which the transformation acts upon somehow have different configuration to the reader from which they come.  Hence the crash.

However I can't see what the code is doing wrong to bring about this condition.

SET DECIMAL=DOT.

DATA LIST notable LIST /text (A24) Y X1 X2 *.
BEGIN DATA.
V00276601                            0.00 90.00 10.00
V00292909                           10.00 30.00 10.00
V00291204                           20.00 20.00 30.00
V00300070                            0.00 90.00 0.00
END DATA.



  • SPLIT FILE BY A.


REGRESSION
        /VARIABLES= Y
        /DEPENDENT= X1 X2
        /METHOD=ENTER
        /STATISTICS=COEFF R ANOVA
        /SAVE= PRED RESID.

LIST.

John Darrington <jmd>
Group administrator
Sun 09 Aug 2015 03:17:46 PM UTC, comment #7: 

I finally managed to reproduce this crash.

I have attached the syntax which provokes it, and a stack trace.

It looks like it's in Ben's area of expertise, so I'll wait until he gets back from vacation to see what he says.


(file #34613, file #34614)

John Darrington <jmd>
Group administrator
Tue 04 Aug 2015 07:40:50 PM UTC, comment #6: 

With
===

data list list
 / v1 to v4.
begin data.
 12 12 89 1
 34 12 80 2
 56 12 77 3
 78 12 73 4
 90 91 85 5
 37 97 85 6
 52 16  82 7
 12 35 79 8
 26 78 76 9
 29 13 71 10
end data.
list.

REGRESSION
   /VARIABLES= v1 v2 v3
   /DEPENDENT=    v4
   /STATISTICS=COEFF R ANOVA
     /SAVE= RESID.
============
I can't provoke a crash with the GTK3 version of PSPP on MSWindows either.

Harry Thijssen <unknown_one>
Tue 04 Aug 2015 04:58:37 AM UTC, comment #5: 

If you click "Paste" instead of "OK" then you will be able to see the syntax before it is run.  Also, if you check the pspp.jnl file, it should give a log of all commands that were run.

I still can't reproduce the problem - perhaps it is something specific to windows.

Until we can get an exact set of circumstances to provoke the problem, it is going to be hard to diagnose.

John Darrington <jmd>
Group administrator
Tue 04 Aug 2015 12:38:40 AM UTC, comment #4: 

I think I might have the same problem, but please bear with me as I'm putting this bug report in for someone else and have no experience with actually running PSPP myself. Not just the software but the concepts themselves are new to me.

OS: Windows 7, 64 bit.
PSPP Version: 0.8.5-g2d71ac.
Problem: When running a linear regression on a dataset, we get no problems when saving the predicted values, but trying to save residuals (with or without the predicted values as well) causes PSPP to crash.

Because of the crash, I don't know the syntax of the command causing the problem, but the command that does work, i.e. when saving with the predicted value but not the residuals, apparently is:

REGRESSION
   /VARABLES= C10 C2 C3
   /DEPENDENT=    C13
   /STATISTICS=COEFF R ANOVA
   /SAVE= PRED.

Presumably the command that causes the crash would have:
    /SAVE= RESID.
or something similar instead.

Phil Hoenig <phoenig>
Mon 20 Apr 2015 07:11:56 AM UTC, comment #3: 

I still cannot reproduce the problem.

Please send the exact syntax you are running, so that we can reproduce the problem, investigate it and fix it.  Thank you.

John Darrington <jmd>
Group administrator
Sun 19 Apr 2015 03:44:42 PM UTC, comment #2: 

I'm afraid the video is too poor a quality for me to see anything.

The regression dialog box (as with all dialog boxes in PSPP/SPSS) emits the syntax if you click the "Paste" button.

Can you send this emitted syntax, and the .sav file too.

John Darrington <jmd>
Group administrator
Sun 19 Apr 2015 11:18:03 AM UTC, comment #1: 

Can you please post a sample of the syntax which causes this and if possible the dataset too.

Thank you.

John Darrington <jmd>
Group administrator
Sun 19 Apr 2015 07:26:36 AM UTC, original submission:  

Hi,
4/19/2015. I just installed
PSPP_0.8.4_2015-03-27_32bits version after also trying the 64 bit one (same release date)....

I'm using Windows 8.1, 64 bit (so tried both installs).
I just happened to need to run linear regression for something and Murphy's law, guess what crashes each time I specify additional "statistics" in the linear regression window. There are 2 by default and I'm asking for the others, and then it crashes. If I don't ask for the other two options (one is the Beta weights, I forget the other checkbox), then it runs fine, but when I ask for these other 2, CRASH! No other info other than the windows "recovery" box that tries to search for a "solution to the problem" :)

Thanks,
Scott

Anonymous

 

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

Attach Files:
   
   
Comment:
   

Attached Files
file #34613:  stacktrace.txt added by jmd (2KiB - text/plain)
file #34614:  crash.sps added by jmd (179B - application/octet-stream)
file #33723:  2-SEU_StuDF_SD_-outliers.sav added by None (250KiB - application/x-spss-sav - I think it might just be this file, as regression is working with options with other files, but I'm not sure why this attached file is crashing the program....)
file #33721:  reg crash.mp4 added by None (513KiB - video/mp4 - Attached is a few seconds video of what happens.... no syntax gets added to the output window, only syntax is from opening the .sav file. Hope this helps)

 

Depends on the following items: None found

Items that depend on this one: None found

 

Carbon-Copy List
  • -email is unavailable- added by blp (Posted a comment)
  • -email is unavailable- added by unknown_one (Posted a comment)
  • -email is unavailable- added by phoenig (Posted a comment)
  • -email is unavailable- added by jmd (Posted a comment)
  •  

    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 logged-in users can vote.

     

    Follow 10 latest changes.

    Date Changed by Updated Field Previous Value => Replaced by
    2016-01-04 blp StatusConfirmed Fixed
        Assigned toNone blp
        Open/ClosedOpen Closed
    2015-08-09 jmd Attached File- Added stacktrace.txt, #34613
        Attached File- Added crash.sps, #34614
        Severity5 - Average 7 - Major
        StatusCannot Reproduce Confirmed
    2015-04-19 None Attached File- Added 2-SEU_StuDF_SD_-outliers.sav, #33723
    2015-04-19 jmd StatusNone Cannot Reproduce
    2015-04-19 None Attached File- Added reg crash.mp4, #33721

    Back to the top

    Powered by Savane 3.13-f8d8.
    Corresponding source code