>= 12204 broken on PPC

[m1b]

While debugging #4090, I've discovered that [12204] causes a crash on ppc (looks like something jumped to null) regardless of whether it's a debug build from xpjets or a local release build. [12203] does not exhibit this problem.

I understand if this gets closed as wontfix since ppc support is being removed for 2.30, but I wanted to report it for compleness's sake.

Exception Type:  EXC_BAD_ACCESS (SIGBUS)
Exception Codes: KERN_PROTECTION_FAILURE at 0x0000000000000000
Crashed Thread:  1

Thread 1 Crashed:
0   org.m0k.transmission          	0x00066e88 tr_buildPath + 296
1   org.m0k.transmission          	0x00078830 tr_sessionInitImpl + 376
2   org.m0k.transmission          	0x000922e8 readFromPipe + 340
3   org.m0k.transmission          	0x000d9c70 event_base_loop + 2432
4   org.m0k.transmission          	0x000d9f20 event_base_dispatch + 16
5   org.m0k.transmission          	0x000920d8 libeventThreadFunc + 188
6   org.m0k.transmission          	0x00081f38 ThreadFunc + 32
7   libSystem.B.dylib             	0x9284b0d0 _pthread_start + 316

[livings124]

Can you post the whole crash log?

[m1b]

​http://pastebin.com/XHTEXmER

[m1b]

Just tried upgrading to [12261] in an effort to verify the fix to #4153, but I still get this crash on launch. Any chance of getting this one fixed? Thanks.

[jordan]

m1b, is there any way to run your build in gdb s.t. you can get a gdb backtrace? That assertion crash report isn't enough information to fix the crash with.

[m1b]

I can try -- I don't have a toolchain on this machine, so I need to go find one and get gdb installed. Could you let me know precisely what you want me to do? I am not particularly gdb literate, so step by step would be appreciated.

[jordan]

Well, if you're more comfortable with another tool that would be fine. What I'm wanting to find out is what arguments were passed into tr_buildPath() and what it's trying to do when it crashes... the "+296" in the crash report above doesn't really help much. :(

[m1b]

I can offer the following:

Program received signal EXC_BAD_ACCESS, Could not access memory.
Reason: KERN_PROTECTION_FAILURE at address: 0x00000000

0x00096b10 in tr_buildPath (first_element=0x0) at utils.c:620

#0  0x00096b10 in tr_buildPath (first_element=0x0) at utils.c:620
#1  0x000a9c64 in tr_sessionInitImpl (vdata=0xbfffeec0) at session.c:685
#2  0x000d3d64 in readFromPipe (fd=4, eventType=2, veh=0x383c40) at trevent.c:192
#3  0x00137898 in event_persist_closure (base=0x387cc0, ev=0x387fd0) at event.c:1238
#4  0x00137330 in event_process_active_single_queue (base=0x387cc0, activeq=0x387e40) at event.c:1282
#5  0x00137b24 in event_process_active (base=0x387cc0) at event.c:1354
#6  0x001384cc in event_base_loop (base=0x387cc0, flags=0) at event.c:1550
#7  0x00137c10 in event_base_dispatch (event_base=0x387cc0) at event.c:1382
#8  0x000d40b0 in libeventThreadFunc (veh=0x383c40) at trevent.c:248
#9  0x000bd1a0 in ThreadFunc (_t=0x387c90) at platform.c:118
#10 0x9284b0d4 in _pthread_start ()

This is with [12294]

NB that I have to build on a different machine than the one gdb is on, so this is all fairly cumbersome.

[m1b]

Adding ​http://pastebin.com/aiF2eab0 for completeness.

[livings124]

m1b: Are you still seeing this issue on current nightlies? No one on Intel (or another OS) has reported this, so it is quite possible that this is a build issue for PPC.

[m1b]

livings124: yes, i just repro'ed it with 12359. jordan and i did some remote debugging a week ago and were able to isolate the rough location of where the relevant var was getting clobbered, but we've had a v.hard time reconnecting to finish the bug hunt.

[livings124]

Based on further discussions, this seems to be a build issue with PPC with the current configuration. Since we're dropping PPC, there's really no reason to pursue this.