issue to open SPM8 standalone and SPM12 standalone with the latest singularity image

[manuegrx]

Hi,

I use the last singularity image (casa-dev-5.0-4.sif ) and I cannot open SPM8Standalone or SPM12Standalone.

For SPM12 standalone I had the following error:

------------------------------------------------------------------------
       Segmentation violation detected at Thu Jul  8 15:24:38 2021
------------------------------------------------------------------------
Configuration:
  MATLAB Version:   7.10.0.499 (R2010a)
  MATLAB License:   [unknown]([url](url))
  Operating System: Linux 4.4.0-186-generic #216-Ubuntu SMP Wed Jul 1 05:34:05 UTC 2020 x86_64
  GNU C Library:    2.27 stable
  Window System:    The X.Org Foundation (11804000), display :0
  Current Visual:   0x21 (class 4, depth 24)
  Processor ID:     x86 Family 6 Model 13 Stepping 7, GenuineIntel
  Virtual Machine:  Java 1.6.0_12-b04 with Sun Microsystems Inc. Java HotSpot(TM) 64-Bit Server VM mixed mode
  Default Encoding:  UTF-8
Fault Count: 1
Register State:
  rax = 0000000000000000   rbx = 00007fe6a0022230
  rcx = 0000000000000000   rdx = 00007fe6a0000080
  rbp = 0000000000000000   rsi = 000000003e07c725
  rdi = 0000000000000000   rsp = 00007fe6a5216da8
   r8 = 0000000000000064    r9 = 0000000000000060
  r10 = 00007fe6a0000080   r11 = 00007fe6a0000080
  r12 = 0000000000000000   r13 = 0000000000000000
  r14 = 00007fe6a52170c8   r15 = 0000000000000000
  rip = 00007fe6b437e426   flg = 0000000000010293
Stack Trace:
If this problem is reproducible, please submit a Service Request via:
    http://www.mathworks.com/support/contact_us/ts/help_request_1.html
A technical support engineer might contact you with further information.
Thank you for your help.
Warning: 1 invisible figure(s) exist at MCR termination.

For SPM8 standalone it is also a " Segmentation violation" error but it is quite long .

I do not have theses issue when opening the same SPM standalone in my computer without casa.
It seems that I can open SPM12 via matlab without issue.

Is there any change in the latest image that can explain this issue ?
Am I the only one to have trouble ?

Emmanuelle

[Hboni]

I just tried to open my SPM12 standalone from casa-distro with the casa-dev-5.0.4 image, and I get the same error.

But I also tried with another version of an SPM12 standalone which contains CAT12, and it opens well. Both versions do not use the same Matlab Runtime (MCR).

[denisri]

Have you checked that the problem does not happen using the casa-dev-5.0-3.sif image ? I'm not sure to understand everything: if the matlab MCR has changed, it means that you are using a different installation of SPM ?

[denisri]

I am only using SPM standalone for the normalization processes of Morphologist in brainvisa, and I still could run it without error on my laptop (using a rather old release of SPM12 standalone).

[Hboni]

I made more tests with SPM12 and SPM8 standalone with different images version. I tried to use a brainvisa coregister process with both versions. I also tried to open the interface of SPM inside a bv bash :

	5.0-4	5.0-3	5.0-1
SPM8 Coreg	error	error	OK
SPM12 Coreg	OK	OK	OK
SPM8 GUI	error	error	OK
SPM12 GUI	error	error	OK

Both SPM8 and SPM12 standalone use an mcr dedicated, but with the exact same version (v713). I tried to use the mcr located in SPM12 folder to launch SPM8 (and the SPM8 mcr for SPM12). And I get the exact same results as in the table.
Furthermore I get a more verbose error with SPM8:

/ __)(  _ \(  \/  )                                           
\__ \ )___/ )    (   Statistical Parametric Mapping           
(___/(__)  (_/\/\_)  SPM8 - http://www.fil.ion.ucl.ac.uk/spm/


------------------------------------------------------------------------
       Segmentation violation detected at Thu Jul 15 14:46:17 2021
------------------------------------------------------------------------

Configuration:
  MATLAB Version:   7.10.0.499 (R2010a)
  MATLAB License:   unknown
  Operating System: Linux 4.13.0-43-generic #48~16.04.1-Ubuntu SMP Thu May 17 12:56:46 UTC 2018 x86_64
  GNU C Library:    2.27 stable
  Window System:    The X.Org Foundation (11804000), display :0
  Current Visual:   0x21 (class 4, depth 24)
  Processor ID:     x86 Family 6 Model 14 Stepping 9, GenuineIntel
  Virtual Machine:  Java 1.6.0_12-b04 with Sun Microsystems Inc. Java HotSpot(TM) 64-Bit Server VM mixed mode
  Default Encoding:  UTF-8

Fault Count: 1

Register State:
  rax = 0000000000000000   rbx = 00007f632404aa40
  rcx = 0000000000000000   rdx = 0000000000000000
  rbp = 0000000000000000   rsi = 000000003e07c725
  rdi = 0000000000000000   rsp = 00007f6329adabc8
   r8 = 0000000000000001    r9 = 0000000000000000
  r10 = 00007f63240008d0   r11 = 00007f6338d48514
  r12 = 0000000000000000   r13 = 0000000000000000
  r14 = 00007f6329adaee8   r15 = 0000000000000000
  rip = 00007f6338d274e1   flg = 0000000000010283

Stack Trace:

If this problem is reproducible, please submit a Service Request via:
    http://www.mathworks.com/support/contact_us/ts/help_request_1.html

A technical support engineer might contact you with further information.

Thank you for your help.


------------------------------------------------------------------------
       Segmentation violation detected at Thu Jul 15 14:46:17 2021
------------------------------------------------------------------------

Configuration:
  MATLAB Version:   7.10.0.499 (R2010a)
  MATLAB License:   unknown
  Operating System: Linux 4.13.0-43-generic #48~16.04.1-Ubuntu SMP Thu May 17 12:56:46 UTC 2018 x86_64
  GNU C Library:    2.27 stable
  Window System:    The X.Org Foundation (11804000), display :0
  Current Visual:   0x21 (class 4, depth 24)
  Processor ID:     x86 Family 6 Model 14 Stepping 9, GenuineIntel
  Virtual Machine:  Java 1.6.0_12-b04 with Sun Microsystems Inc. Java HotSpot(TM) 64-Bit Server VM mixed mode
  Default Encoding:  UTF-8

Fault Count: 2

Register State:
  rax = 00007f63399c2700   rbx = 0000000000000000
  rcx = 0000000000000001   rdx = 00007f6329add430
  rbp = 00007f6329add110   rsi = 00007f6327724a70
  rdi = 0000000000000001   rsp = 00007f6329add0d0
   r8 = 00007f6325d30d30    r9 = 0000000000000001
  r10 = 0000000000000000   r11 = 0000000000000001
  r12 = 00007f6329add2b0   r13 = 0000000000000001
  r14 = 0000000000000002   r15 = 0000000000000001
  rip = 00007f6333051954   flg = 0000000000010206

Stack Trace:
  [0] libmwm_interpreter.so:inAssignInputArgsToFormals(int, mxArray_tag**, Mfh_mp*, int)(133, 0x7f6329add370, 0, 0x7f6327724a70) + 84 bytes
  [1] libmwm_interpreter.so:inRunMfile(int, mxArray_tag**, int, mxArray_tag**, Mfh_mp*, inWorkSpace_tag*)(0x7f6327724a70, 0x13869f298, 0x7f6327548a50, 0x7f63399c2700) + 1404 bytes
  [2] libmwm_dispatcher.so:Mfh_file::dispatch_fh(int, mxArray_tag**, int, mxArray_tag**)(0x7f6329add500, 0x7f6329add6f0, 0x7f6329add870, 0x7f6327548e10) + 330 bytes
  [3] libmwmcos.so:mcos::MScopedMethodInfo::invoke(mcos::COSClient const*, int&, mcos::COSValue*, int, mcos::COSValue const*) const(0x7f63399c2170, 0x7f6329add6e0, 0x7f6329add6b0, 0x7f6329add6c0) + 349 bytes
  [4] libmwmcos.so:mcos::MDestructorMI::invoke(mcos::COSClient const*, int&, mcos::COSValue*, int, mcos::COSValue const*) const(0x7f6329add890, 0x7f6329add8b0, 0x7f6329add920, 0x7f632755e690) + 506 bytes
  [5] libmwmcos.so:mcos::COSClassI::callDestructor(mcos::COSInterfacePtr)(0x7f6329addaf0, 0x7f6327547750, 0x7f632754f080, 0x7f6329adda00) + 483 bytes
  [6] libmwmcos.so:mcos::COSClassI::destroyInstance(mcos::COSInterfacePtr)(0x7f6329add990, 0x7f632f7b9e44, 0x7f6329adda40, 0x7f632f7a912c) + 110 bytes
  [7] libmwmcos.so:mcos::COSInterfaceI::destroyUnconditional()(0x7f6329addb90, 0x7f6329addbc8, 0x7f632754f220, 0x7f62c3098280) + 1826 bytes
  [8] libmwmcos.so:mcos::COSInterfaceI::destroyRef()(0x7f62c3096a20, 0, 0x7f6329addbd0, 0x7f632f83e075) + 134 bytes
  [9] libmwmcos.so:mcos::COSInterfacePtr::release()(0x7f62c3098280, 0x7f632754f080, 0x7f6329addc10, 0x7f63381b4317) + 37 bytes
  [10] libmwmcos.so:omDestroyElementListener(void*, mxArray_tag const*)(2, 0x7f62c3098280, 17, 0x7f6327220b60) + 69 bytes
  [11] libmx.so:0x00007f63381b4317(0x7f62c3098280, 0x7f63274ab020, 0x7f6329addc60, 0x7f63381b4c4a)
  [12] libmx.so:0x00007f63381b38c0(0x7f6329addc60, 0x7f6327220bb8, 0x7f63274ab020, 0x7f6327220b60)
  [13] libmx.so:mxDestroyArray~(0x7f6329addca0, 0, 0, 0x7f63274ab020) + 74 bytes
  [14] libmwm_interpreter.so:inResetWS(inWorkSpace_tag*, int)(0xffffffff0000003f, 0x7f6329addd00, 0x274b1900, 0x7f63274ab020) + 242 bytes
  [15] libmwm_interpreter.so:inMfileEpilogFinalizer(opcodes, void*, opcodes*, int*)(0x7f6329addd90, 0x7f632778b690, 0x7f6329addf30, 0x7f632745c2d0) + 766 bytes
  [16] libmwm_interpreter.so:inRunFinalizers(opcodes, opcodes*, int*)(0x7f6329adddd0, 0x7f6329addff0, 0, 0x7f6329ade240) + 55 bytes
  [17] libmwm_interpreter.so:inRunMfile(int, mxArray_tag**, int, mxArray_tag**, Mfh_mp*, inWorkSpace_tag*)(0x7f6329ade240, 0x200000563, 0x7f632745c2d0, 0x7f6329adefc0) + 2720 bytes
  [18] libmwm_dispatcher.so:Mfh_file::dispatch_fh(int, mxArray_tag**, int, mxArray_tag**)(0x200000002, 0x7f63273f31c8 "hgload", 0x7f6329ad0000, 0x7f6327404580) + 330 bytes
  [19] libmwm_interpreter.so:inDispatchFromStack(int, char const*, int, int)(0x7f6329ade5ec, 2, 0x7f63273f31c8 "hgload", 0) + 1066 bytes

[denisri]

So,

• SPM12 standalone coreg (and probably normalise) do work with all images, that's probably why we didn't run into this problem earlier.
• the problem did not appear in casa-dev-5.0-4 but in 5.0-3 or 5.0-2.

Now after testing a bit, I will add even more confusion: I have 2 machines (a desktop and a laptop), both running Ubuntu 18.04 (with updates done), with SPM12-standalone-7771 installed (same release on both machines). Casa-distro updated to image 5.0-4.

• on my desktop machine, SPM12 GUI starts without error in casa-distro.
• Same when run natively on the host system.
• on my laptop, I get the same crash as you get in casa-distro.
• But... I also get the crash outside of casa-distro container on the host system...

So would it be linked to a user config somewhere (path...) ?

[DimitriPapadopoulos]

MCR v713 is the version shipping with Matlab R2010a.

• MATLAB Runtime
• What is the version of the MATLAB Compiler Runtime (MCR) that corresponds to the version of MATLAB Compiler I am using?

[DimitriPapadopoulos]

Perhaps ReproNim/neurodocker#358 will help.

[DimitriPapadopoulos]

But then, do I need to update the MCR used by the standalone SPM outside of Singularity? Why hasn't this ever been reported outside Singularity here at NS?

[gllmflndn]

Hello 😀

I couldn't follow all your tests and I don't know what casa-distro is but there is a good chance all of the issues will disappear when using a more recent version of the MCR. R2010a was used consistently to prevent having to install a new MCR from time to time (or having to keep several versions in parallel) but it is just very dated now. I use R2019b in this Singularity definition file with no known issue.
Standalone SPM12 with R2019b is not an official release but I could try to make that happen (at least for Linux) soon if that was helpful here.

[DimitriPapadopoulos]

Hi Guillaume! So to be clear, you recommend we deploy the SPM8 and SPM12 standalone versions with the MCR v97 from Matlab R2019b? Although this ticket is about a Singularity deployment, I am personally interested in the deployment of SPM directly on Ubuntu 18.04 and 20.04 workstations, without Singularity.

If you envision a new SPM12 standalone release, why not switch directly to R2021a?

[gllmflndn]

Yes, the idea would be to release a new SPM12 standalone version (r7771) relying on the most recently available version of the MATLAB Runtime (probably waiting a few weeks for R2021b, even if ideally I was waiting for R2022a). From then on, would you prefer if new releases of SPM standalone were using the same version of the MATLAB Runtime or if they were using the most recently available one at each given time (or both)?
Also make sure to set an SPM_HTML_BROWSER environment variable to 0 before starting SPM as it is known otherwise to be the cause of a crash.
This plan does not concern SPM8 though - is this still of any use to you?

[DimitriPapadopoulos]

If a new MCR is required to fix bugs or crashes, then by all means please do upgrade the MCR - that's more important than saving a few GB of disk space. On the other hand, I suggest using the same MCR as long as it works and has no security issues - every sub-system upgrade comes with its own (minimal) risks of problems.

Good to learn about this SPM_HTML_BROWSER trick for r7771.

I tried to erase SPM5, but a couple users still need it in some scripts! As you can imagine, we have even more users still in need of SPM8 here and there. One last update of SPM8 based on the latest version of Matlab would be useful. It should buy us enough time (years) to phase out SPM8 completely...

[DimitriPapadopoulos]

Has this issue been fixed in the Singularity image? How?