[R-SIG-Mac] R segfault at startup, causing R crash
Kevin Ushey
kevinushey at gmail.com
Sun Nov 6 06:58:45 CET 2016
After playing around with R.app for a bit, I was able to get myself
into a state where R.app was either dying or hanging on startup. I did
the following:
1. Open R.app,
2. Open a new R document (using the icon in the top menu bar --
'Create a new, empty document in the editor' ),
3. Write and execute some code -- in my case, 'x <- 1; print(x)',
executed with Cmd + Enter,
4. (optionally) Repeat steps 2 and 3, creating new documents and
executing more code,
5. Focus a different application,
6. Attempt to open an R document (using the menu bar again --
'Source script or load data in R',
7. Observe that R.app is now hanging.
Note that if you run these repro steps, you may see R segfault / die
on startup (although in some cases all was fine). Note that step 5 is
necessary (IIUC) to force R.app to attempt to save the application
state.
It looks like R.app stores session state in the folder
~/Library/Saved Application State/org.R-project.R.savedState
My hypothesis at this point is that this saved state is somehow
getting corrupted. Removing that folder seems to resolve any startup
crashes I see.
As for the 'Open File' dialog hang, here's what I believe are the
relevant bits of an lldb stack trace I see:
* frame #0: 0x00007fffce2a7cba libsystem_kernel.dylib`__psynch_mutexwait + 10
frame #1: 0x00007fffce390f6a
libsystem_pthread.dylib`_pthread_mutex_lock_wait + 100
frame #2: 0x00007fffc573faee
FinderKit`+[FI_TFloatingInputWindowController singleton] + 34
frame #3: 0x00007fffc57edd5b FinderKit`-[FI_TBrowserTableView
inputContext] + 29
frame #4: 0x00007fffb6d397dc AppKit`+[NSTextInputContext
currentInputContext] + 145
frame #5: 0x00007fffb6e7a272 AppKit`+[_NSAutomaticFocusRing
setActiveFirstResponderChanged] + 37
frame #6: 0x00007fffb6e7a1d9
AppKit`___NSPostActiveFirstResponderChanged_block_invoke + 29
frame #7: 0x00007fffb90cdec7
CoreFoundation`__CFRUNLOOP_IS_CALLING_OUT_TO_AN_OBSERVER_CALLBACK_FUNCTION__
+ 23
frame #8: 0x00007fffb90cde37 CoreFoundation`__CFRunLoopDoObservers + 391
frame #9: 0x00007fffb90aead9 CoreFoundation`__CFRunLoopRun + 873
frame #10: 0x00007fffb90ae514 CoreFoundation`CFRunLoopRunSpecific + 420
frame #11: 0x00007fffbaaabcb2 Foundation`-[NSRunLoop(NSRunLoop)
runMode:beforeDate:] + 277
frame #12: 0x000000010b21f43e R`-[RDocumentWinCtrl
setStatusLineText:](self=0x00006180001e7c00,
_cmd="setStatusLineText:", text=@"print(x, ...)") + 910 at
RDocumentWinCtrl.m:465
frame #13: 0x000000010b21f03d R`-[RDocumentWinCtrl
RDocumentDidResize:](self=0x00006180001e7c00,_cmd="RDocumentDidResize:",
notification=@"NSWindowDidResizeNotification") + 77 at
RDocumentWinCtrl.m:431
with context from frame 12:
frame #12: 0x000000010b21f43e R`-[RDocumentWinCtrl
setStatusLineText:](self=0x00006180001e7c00,
_cmd="setStatusLineText:", text=@"print(x, ...)") + 910 at
RDocumentWinCtrl.m:465
462 [statusLine setStringValue:text];
463 [statusLine setNeedsDisplay:YES];
464 // Run NSDefaultRunLoopMode to allow to update status line
-> 465 [[NSRunLoop currentRunLoop] runMode:NSDefaultRunLoopMode
466
beforeDate:[NSDate distantPast]];
467
468 }
Hopefully this helps...
Best,
Kevin
---
R version 3.3.2 (2016-10-31) -- "Sincere Pumpkin Patch"
Copyright (C) 2016 The R Foundation for Statistical Computing
Platform: x86_64-apple-darwin13.4.0 (64-bit)
macOS Sierra -- Version 10.12.1
On Sat, Nov 5, 2016 at 5:08 PM, Kevin Ushey <kevinushey at gmail.com> wrote:
> It might be worth trying to launch R with lldb attached. You can run
> the following command from the terminal:
>
> /usr/bin/lldb -w -n R
>
> Then, try launching R (or R.app) -- 'lldb' should then detect that R
> has been launched, and you can enter 'c' in the 'lldb' terminal window
> to continue execution. Once R crashes / segfaults, you should
> hopefully get a nice backtrace that may provide some actionable
> information on how the bug could be fixed. The backtrace can normally
> be printed using 'bt' in the lldb console. If lucky, that backtrace
> should show exactly what code is causing the crash, and could give
> some hints on how the crash could be resolved.
>
> If you're not familiar with lldb, http://lldb.llvm.org/tutorial.html
> gives a nice tutorial.
>
> Best,
> Kevin
>
> On Sat, Nov 5, 2016 at 8:55 AM, C W <tmrsg11 at gmail.com> wrote:
>> Hi Bryan,
>>
>> Right after I wrote the email, I restarted R and the problem came back. I
>> guess I jinx-ed my luck. ;)
>>
>> The funny thing is, I did not even have a Rscript, it was just the stand
>> alone R console. I was actually using it as a calculator.
>>
>> I always thought OS X is more compatible with R than Windows, who knows. :(
>>
>> On Sat, Nov 5, 2016 at 11:47 AM, Bryan Hanson <hanson at depauw.edu> wrote:
>>
>>> No problem, glad to try to help.
>>>
>>> There probably isn’t a problem with R.gui itself (when you “run R” on a
>>> Mac by clicking the icon you are causing your computer to actually run the
>>> R.gui which in turn is running R under the hood).
>>>
>>> So if the problem re-occurs, note the name of the most recent file (it
>>> probably was opened when you launched R prior to R faulting). Then find
>>> that file and change the extension (temporarily). Then try to launch R
>>> again and I expect it will not have a problem.
>>>
>>> Bryan
>>>
>>> > On Nov 5, 2016, at 11:43 AM, C W <tmrsg11 at gmail.com> wrote:
>>> >
>>> > Hi Bryan,
>>> >
>>> > I appreciate your detailed explanation. I am not an R expert, I do not
>>> know how to fix the R.gui you mentioned.
>>> >
>>> > I kept on opening and closing R, after a dozen tries, the error went
>>> away temporarily. I still get the error, but it was not every time.
>>> >
>>> > Since then, I have upgraded to R version 3.3.2 (2016-10-31).
>>> >
>>> > I think this problem may come back, please let me know when you find a
>>> solution. Thank you!
>>> >
>>> > On Fri, Nov 4, 2016 at 8:28 PM, Bryan Hanson <hanson at depauw.edu> wrote:
>>> > Sorry, I should add that when you rename the offending file you need to
>>> change its extension, otherwise R will follow the file.
>>> >
>>> > >>> Original Message:
>>> >
>>> > I had something similar earlier this week (also on Sierra). I was in
>>> touch with Simon who suggested it might be a bad workspace image, so you
>>> can try to remove any .RData or .rda files. However, in my case it was not
>>> that (I don’t save workspace images or histories for that matter). But I
>>> removed all of those from my entire computer anyway and the problem
>>> remained (a few were left on shared drives by others who were saving
>>> workspace images).
>>> >
>>> > Ultimately I was able to trace the problem to something being wrong with
>>> the “recent” list. In my case R.gui was trying to open the most recent
>>> file in the list, but something about the link to it was corrupted. I
>>> verified this by running R from the command line and sourcing the file that
>>> seemed to be associated with the crash. It worked fine, so the file was
>>> not the problem.
>>> >
>>> > So, here’s what worked for me: When R.gui is asking you for a
>>> “selection” to close down, see if the File -> Open Recent… option can still
>>> be selected, and if so, go all the way down to the clear menu item. Once
>>> this list of recently opened files was empty, R.gui ran fine. In my case
>>> not all files would clear from the list. There was one left, and it was
>>> the one apparently causing the problem. I renamed it so R.gui couldn’t use
>>> it/find it, and then the problem was gone. Again, there was nothing wrong
>>> with the file, there was something wrong with the “remembering”.
>>> >
>>> > A few days later, I had to repeat this process when the problem
>>> re-occurred. There may be something genuinely conflicted between Sierra
>>> and R.gui, only time will tell. If it proves to be a widespread problem
>>> more observations should help narrow it down. But this is what I
>>> experienced and how I fixed it.
>>> >
>>> > Bryan
>>> >
>>> >
>>> > > On Nov 4, 2016, at 7:31 PM, C W <tmrsg11 at gmail.com> wrote:
>>> > >
>>> > > Dear R list,
>>> > >
>>> > > Every time I start R, the following error message come up:
>>> > >
>>> > > *** caught segfault ***
>>> > > address 0x0, cause 'unknown'
>>> > >
>>> > > Possible actions:
>>> > > 1: abort (with core dump, if enabled)
>>> > > 2: normal R exit
>>> > > 3: exit R without saving workspace
>>> > > 4: exit R saving workspace
>>> > >>
>>> > > Selection:
>>> > >
>>> > >
>>> > > Sometimes I get this error message instead,
>>> > >
>>> > > 2016-11-04 16:50:08.446 R[23372:326103] -[NSDrawerWindow
>>> > > setRuleThickness:]: unrecognized selector sent to instance
>>> 0x7fabc94a74b0
>>> > > 2016-11-04 16:50:08.454 R[23372:326103] *** RController: caught ObjC
>>> > > exception while processing system events. Update to the latest GUI
>>> version
>>> > > and consider reporting this properly (see FAQ) if it persists and is
>>> not
>>> > > known.
>>> > > *** reason: -[NSDrawerWindow setRuleThickness:]: unrecognized selector
>>> sent
>>> > > to instance 0x7fabc94a74b0
>>> > > *** name: NSInvalidArgumentException, info: (null)
>>> > > *** Version: R 3.3.2 (71607) R.app R 3.3.2 GUI 1.68 Mavericks build
>>> > > Consider saving your work soon in case this develops into a problem.
>>> > > Here's the sessionInfo(),
>>> > >
>>> > >> sessionInfo()
>>> > > R version 3.3.2 (2016-10-31)
>>> > > Platform: x86_64-apple-darwin13.4.0 (64-bit)
>>> > > Running under: macOS Sierra 10.12.1
>>> > >
>>> > > locale:
>>> > > [1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8
>>> > >
>>> > > attached base packages:
>>> > > [1] stats graphics grDevices utils datasets methods
>>> > > [7] base
>>> > >
>>> > >
>>> > > I would appreciate any help, thank you!
>>> > >
>>> > > [[alternative HTML version deleted]]
>>> > >
>>> > > _______________________________________________
>>> > > R-SIG-Mac mailing list
>>> > > R-SIG-Mac at r-project.org
>>> > > https://stat.ethz.ch/mailman/listinfo/r-sig-mac
>>> >
>>> >
>>>
>>>
>>
>> [[alternative HTML version deleted]]
>>
>> _______________________________________________
>> R-SIG-Mac mailing list
>> R-SIG-Mac at r-project.org
>> https://stat.ethz.ch/mailman/listinfo/r-sig-mac
More information about the R-SIG-Mac
mailing list