[Vm-dev] Pharo 6 Segmentation fault on Travis OS X

[Ben Coman]

On Fri, Apr 21, 2017 at 8:11 AM, Juraj Kubelka <juraj.kubelka at icloud.com> wrote:
>
> Hi,
>
> I have noticed that Pharo 6 VM is quite unstable on OS X. I do not face it on my desktop, but my Travis CI builds are mostly broken. See the example here: https://travis-ci.org/JurajKubelka/PharoSprint/jobs/224161107#L258
>
> Segmentation fault Thu Apr 20 23:31:54 2017
> VM: 201702221539 https://github.com/pharo-project/pharo-vm.git $ Date: Wed Feb 22 16:39:40 2017 +0100 $
> Plugins: 201702221539 https://github.com/pharo-project/pharo-vm.git $
> C stack backtrace & registers:
> eax 0x00000000 ebx 0x044f9014 ecx 0x044f9014 edx 0x00000000
> edi 0x044f7a7d esi 0x044f7a7d ebp 0xbffabae8 esp 0xbffabaa0
> eip 0x000689e6
> 0   Pharo                               0x000689e6 bytecodePCForstartBcpcin + 195
> 1   Pharo                               0x000961be reportStackState + 706
>
>
>
> The same build task is stable on Windows and Linux.
>
> If some of the builds are green, it is because I have restarted the job, typically several times.
>
> There is also this issue at the beginning of the log: https://travis-ci.org/JurajKubelka/PharoSprint/jobs/224161107#L84
>
>
> 62.56sDownloading the latest pharoVM:
> http://files.pharo.org/get-files/60/pharo-mac-stable.zip
> pharo-vm/Pharo.app/Contents/MacOS/Pharo
> Downloading PharoV60.sources:
> http://files.pharo.org/get-files/60/sources.zip
> Creating starter scripts pharo and pharo-ui
> 2017-04-20 23:30:25.180 Pharo[2532:5750] 23:30:25.180 ERROR:    312: error -66680
> 2017-04-20 23:30:25.182 Pharo[2532:5750] 23:30:25.182 ERROR:    >aq> 1630: failed (-66680); will stop (5512/0 frames)
> 2017-04-20 23:30:25.184 Pharo[2532:5750] 23:30:25.184 ERROR:    312: error -66680
> 2017-04-20 23:30:25.184 Pharo[2532:5750] 23:30:25.184 ERROR:    >aq> 1630: failed (-66680); will stop (33075/0 frames)
>
> Any idea how can it be improved? Or how can I debug it in order to provide more information?
>
> Thanks!
> Juraj
>
>
>

https://travis-ci.org/JurajKubelka/PharoSprint/jobs/224161107#L297-L299
Smalltalk stack dump:
0xbffb7ae8 I SessionManager>snapshot:andQuit: 0x5a26fd0: a(n) SessionManager
0xbffb7b18 I [] in SmalltalkImage>snapshot:andQuit: 0x4fc3410: a(n)
SmalltalkImage

Maybe it is be the Compactor bug,
http://forum.world.st/Crashes-on-snapshot-with-the-new-compactor-td4939959.html#a4942434

introduced I guess per this commit...
    Date:   Sat Feb 18 07:43:47 2017 -0800
    CogVM source as per VMMaker.oscog-eem.2134
    Move the SpurPlanningCompactor to tip.

and fixed
    Date:   Mon Apr 10 19:37:04 2017 +0200
    SpurPlanningCompactor:
    Fix a bug in resetting the free chunk used for the
firstUnusedFieldsSpace after
    non-final pasxses (i.e. on snapshot).  The old code didn't check to see if a
    free chunk was actually found(!!).

where that VM http://files.pharo.org/get-files/60/pharo-mac-stable.zip
falls in between
   VM: 201702221539 [...]  Date: Wed Feb 22 [...]

Try with http://files.pharo.org/get-files/60/pharo-mac-latest.zip

@Esteban, when you're back from vacation, probably good to bump stable prior
doing it again(?) at Pharo 6 release.

cheers -ben