Repository navigation
Mac OS X: Occasional SIGSEGV on mouse (Magic Trackpad) move #5
Description
Process: solvespace [30608]
Path: /Applications/solvespace.app/Contents/MacOS/solvespace
Identifier: solvespace
Version: 2.1 (2.1)
Code Type: X86-64 (Native)
Parent Process: ??? [1]
Responsible: solvespace [30608]
User ID: 502
Date/Time: 2016-05-29 00:24:12.402 +0100
OS Version: Mac OS X 10.11.5 (15F34)
Report Version: 11
Anonymous UUID: 77A06C71-7219-FFF3-7683-72B79C8D32EE
Sleep/Wake UUID: 9B50071F-8246-4CC9-B847-AF35F9AB8A65
Time Awake Since Boot: 140000 seconds
Time Since Wake: 2000 seconds
System Integrity Protection: enabled
Crashed Thread: 0 Dispatch queue: com.apple.main-thread
Exception Type: EXC_BAD_ACCESS (SIGSEGV)
Exception Codes: KERN_INVALID_ADDRESS at 0x0000002000000020
Exception Note: EXC_CORPSE_NOTIFY
VM Regions Near 0x2000000020:
CoreAnimation 000000098dd22000-00000009913a1000 [ 54.5M] rw-/rwx SM=PRV
-->
STACK GUARD 0000700000083000-0000700000084000 [ 4K] ---/rwx SM=NUL stack guard for thread 1
Global Trace Buffer (reverse chronological seconds):
464.718595 libextension.dylib 0x00007fff91303533 tearing down extension request for pid 30645
Thread 0 Crashed:: Dispatch queue: com.apple.main-thread
0 ??? 0x0000002000000020 0 + 137438953504
1 solvespace 0x000000010a460376 SolveSpace::TextWindow::MouseEvent(bool, bool, double, double) + 566
2 solvespace 0x000000010a3ddd44 -[TextWindowView mouseMoved:] + 100
3 com.apple.AppKit 0x00007fff937131e4 -[NSTrackingArea mouseMoved:] + 152
4 com.apple.AppKit 0x00007fff93c99bd1 -[NSWindow _handleMouseMovedEvent:] + 266
5 com.apple.AppKit 0x00007fff93c9ab15 -[NSWindow _reallySendEvent:isDelayedEvent:] + 2108
6 com.apple.AppKit 0x00007fff936d9539 -[NSWindow sendEvent:] + 517
7 com.apple.AppKit 0x00007fff936594db -[NSApplication sendEvent:] + 1167
8 com.apple.AppKit 0x00007fff934c0df2 -[NSApplication run] + 796
9 solvespace 0x000000010a3df458 main + 296
10 solvespace 0x000000010a3c2e54 start + 52
Activity
whitequark commentedon May 28, 2016
I don't have any Apple hardware at all. Is this related just to pointer movement? Can it be reproduced somehow?
michthom commentedon May 29, 2016
I've not found a reproducible scenario - might be tied to the multi-finger gestures on the trackpad - I will see if using a 'regular' mouse is less prone to tickling the bug.
whitequark commentedon May 29, 2016
Any chance you could run SolveSpace under a debugger? It should have all debugging symbols, and give a far more informative backtrace, which would confirm some ideas I have about the cause of this bug.
michthom commentedon May 29, 2016
I'll see what I can do - I have Xcode installed, so should have LLDB - but no clue how to operate it! Will try reading this first: https://developer.apple.com/library/mac/documentation/IDEs/Conceptual/gdb_to_lldb_transition_guide/document/lldb-terminal-workflow-tutorial.html#//apple_ref/doc/uid/TP40012917-CH4-SW1
michthom commentedon May 29, 2016
michthom$ lldb /Applications/solvespace.app/Contents/MacOS/solvespace
(lldb) target stop-hook add
Enter your stop hook command(s). Type 'DONE' to end.
Stop hook #1 added.
(lldb) process launch
michthom commentedon Jun 1, 2016
Process 38892 launched: '/Applications/solvespace.app/Contents/MacOS/solvespace' (x86_64)
failed to look up item 00000000, searched 27 items
oops at line 302, file /Users/travis/build/solvespace/solvespace/src/dsc.h
Process 38892 exited with status = 255 (0x000000ff)
whitequark commentedon Jun 11, 2016
@michthom Unfortunately that still doesn't have a backtrace...
whitequark commentedon Jun 11, 2016
This apparently can be grabbed using the
thread backtracelldb command.michthom commentedon Jun 13, 2016
Thanks - I'll see if I can get more details with this. i had thought that 'bt' would generate a backtrace, but the stop hook seems to be ignored when the app dies - perhaps it only fires on a breakpoint being hit?
I'm now trying to reproduce the problem and trace with:
michthom$ lldb /Applications/solvespace.app/Contents/MacOS/solvespace
(lldb) target stop-hook add
Enter your stop hook command(s). Type 'DONE' to end.
Stop hook #1 added.
(lldb) process launch
^C
Process 4979 stopped
(lldb) process handle SIGSEGV --notify true --pass true --stop true
NAME PASS STOP NOTIFY
=========== ===== ===== ======
SIGSEGV true true true
(lldb) process continue
whitequark commentedon Jun 23, 2016
@michthom Sorry, missed your last comment. I now understand your problem--lldb doesn't stop on abort. I think that a more robust way to fix this, rather than messing with hooks, is to simply set a breakpoint on
abort, which would let you grab a backtrace from there.michthom commentedon Jun 25, 2016
Aha.
However, the backtrace that got me here in the first place didn't actually call abort, so I may use both this and the technique in my previous comment.
michthom commentedon Jun 25, 2016
I'm feeling veeeeery stupid at this point, but are the DSYM files included in the 2.1 dmg? I'm looking under /Applications/solvespace.app/Contents/Resources/ and I can't see anything apart from some nibs, icons and a png?
So when I load a core file into lldb I'm not sure where it's going to get the debugging symbols from?
Or have I been dumb and need to grab the source and build it myself to get the symbols to give you any useful feedback?
Update - grabbed the source, built it and I still see no debug symbol files. Do I need to change cmake overrides or something? Sorry for the dumb questions, once I get this working I'll happily write a 'Solvespace debugging for idiots' guide!
whitequark commentedon Jun 26, 2016
I think they should be in the executable. The CI builds are performed with CMAKE_BUILD_TYPE as either Debug or RelWithDebInfo, and the default build type is equivalent to Debug, so you should have the symbols. No idea what dSYM stuff is about.
whitequark commentedon Jun 26, 2016
Oh, looks like dSYM is for stripped symbols. I.e. you take your proprietary executable and separate the debugging symbols into another file, which you save, and ship the executable. They are inapplicable in case of SolveSpace.
1 remaining item
whitequark commentedon Jun 27, 2016
@michthom Ok, I've figured this out. On OS X (only?), you have to build with -DCMAKE_BUILD_TYPE=Debug to get any debug symbols at all.
@lucdesign As I've mentioned above, I do not have any Apple hardware and cannot reproduce this on my own. You can help me fix this by providing a symbolized backtrace, by following the build instructions (replacing
cmake ..withcmake .. -DCMAKE_BUILD_TYPE=Debug), runninglldb build/src/solvespace.app, then in LLDB prompt enteringrunandbtonce it crashes.michthom commentedon Jun 29, 2016
Sorry I've been away on business, will rebuild a debug version and see if I can reproduce soon.
whitequark commentedon Jul 19, 2016
@michthom ping
michthom commentedon Jul 19, 2016
Argh! Sorry - I've been swamped with work. I have a debug version, I have been able to get it to crash and see the backtrace with symbols, but I've no idea what I was doing at the time. I'll have more time to use SolveSpace in the coming weeks, if you're willing to wait, else close this one and I'll reopen once I have some debug info to provide?
whitequark commentedon Jul 20, 2016
@michthom Sure--I'll wait. No particular rush here, just making sure this issue doesn't die out as it seems serious and I've no means to reproduce it myself.
whitequark commentedon Aug 7, 2016
@michthom Any luck with that backtrace? It shouldn't take too long... I would like to get this fixed in 2.2, which is scheduled in 1-2 weeks from now, as there will likely be no release for a long time after that.
michthom commentedon Aug 7, 2016
Sorry I've failed. My mac is with Apple trying to debug a random freeze/crash and I'm currently on holiday, back August 15th so I think you'll have to go without me for now.
Sent from my iPhone
whitequark commentedon Oct 7, 2016
@michthom ping?
michthom commentedon Oct 7, 2016
Sorry - this task has rather slipped my mind. I'm still battling with Apple - they're currently debating replacement of my iMac as it has got worse not better after three trips to the Genius Bar.
I really don't think I can commit to helping out in a reasonable timescale, so if you're waiting for me to help find that crash bug before releasing 2.2, it may be some time off yet...
If on the other hand you have the patience of Job I will pick this up once I have a stable machine and can spend enough time in a debug build of Solvespace to reproduce the issue. Not having a set of steps that triggers it reliably, it'll be me flailing in the dark for a while before it happens again.
whitequark commentedon Oct 8, 2016
@michthom Understood--I wish you luck with your troubles. It's not that I am impatient, this is just normal triage process, making sure that the project is going forward.
whitequark commentedon Oct 10, 2016
Fixed in master and 2.x (via #63).
Merge pull request solvespace#5 from solvespace/master