I used a tarball from the website. I'll try and find out a git/svn url and checkout the head revision of the trunk.
-- Laurent
Hi Laurent,
This doesn't look like the code from trunk. Are you using an old release? As I said, I think this bug is fixed some months ago.
David
On 6 Mar 2013, at 01:12, Laurent Michel <address@hidden> wrote:
David,
I recompiled and here is the trace:
Program received signal SIGSEGV, Segmentation fault. 0xb7a55695 in isEmptyProtocol (aProto=0xb7736f90) at protocol.c:62 62 isEmpty &= (p2->properties->count == 0); (gdb) where #0 0xb7a55695 in isEmptyProtocol (aProto=0xb7736f90) at protocol.c:62 #1 0xb7a5553e in unique_protocol (aProto=0xb7736f90) at protocol.c:118 #2 0xb7a534eb in init_protocols (protocols=0xb7736fb0) at protocol.c:189 #3 0xb7a534da in init_protocols (protocols=0xb773a0c0) at protocol.c:186 #4 0xb7a5328a in objc_init_protocols (protocols=0xb773a0c0) at protocol.c:196 #5 0xb7a48264 in load_category (cat=0xb773a2c0, class=0xb7a64340) at category_loader.c:38 #6 0xb7a47f02 in try_load_category (cat=0xb773a2c0) at category_loader.c:50 #7 0xb7a47e6a in objc_try_load_category (cat=0xb773a2c0) at category_loader.c:63 #8 0xb7a5143f in __objc_exec_class (module=0xb773a398) at loader.c:89 #9 0xb76c55a4 in .objc_load_function () from /home/ldm/work/platform/Clo/queens2/../ORFoundation/libORFoundation.so #10 0xb7fece9b in ?? () from /lib/ld-linux.so.2 #11 0xb7fecf84 in ?? () from /lib/ld-linux.so.2 #12 0xb7fdf20f in ?? () from /lib/ld-linux.so.2
So line 62.
It won't show the variables though:
(gdb) p *p2 $3 = { isa = 0xb7a642f0, name = 0xb771df54 "ORExpr", protocol_list = 0xb7795a18, instance_methods = 0xb7795a20, class_methods = 0xb7795a20, optional_instance_methods = 0xb7795a20, optional_class_methods = 0xb7795a20, properties = 0x0, optional_properties = 0x0 } (gdb) l 57 if (aProto->isa == ObjC2ProtocolClass) 58 { 59 struct objc_protocol2 *p2 = (struct objc_protocol2*)aProto; 60 isEmpty &= (p2->optional_instance_methods->count == 0); 61 isEmpty &= (p2->optional_class_methods->count == 0); 62 isEmpty &= (p2->properties->count == 0); 63 isEmpty &= (p2->optional_properties->count == 0); 64 } 65 return isEmpty; 66 }
So p2->properties is NULL and trying to access the count attribute crashes. This is happening on a protocol of mine (ORExpr) which inherits from other protocols:
@protocol ORExpr <ORConstraint,NSObject,NSCoding> -(id<ORTracker>) tracker; -(ORInt) min; -(ORInt) max; -(BOOL) isConstant; -(BOOL) isVariable;
....
@end
Any ideas?
On Mar 5, 2013, at 7:20 AM, Laurent Michel <address@hidden> wrote:
On Mar 5, 2013, at 4:10 AM, David Chisnall <address@hidden> wrote:
Hi,
How old is your runtime? Without debugging symbols, this looks like it might be a bug that was fixed in May last year. It would really help if you could do a debug build of the runtime and let me know exactly what is failing. If you do ccmake . in your libobjc2 build directory, and change the build type to Debug, you should get the symbols.
libobjc2 was 1.6.1
address@hidden:/usr/local/src/old$ sha1sum libobjc2-1.6.tar.gz 348ab64899b25ecb2bfdfd13ebeeab2d6a2c8988 libobjc2-1.6.tar.gz
total 9196 drwxr-xr-x 4 root root 4096 Mar 4 08:37 GNUstep lrwxrwxrwx 1 root root 23 Mar 4 08:37 libgnustep-base.so -> libgnustep-base.so.1.24 lrwxrwxrwx 1 root root 25 Mar 4 08:37 libgnustep-base.so.1.24 -> libgnustep-base.so.1.24.3 -rwxr-xr-x 1 root root 9105637 Mar 4 08:36 libgnustep-base.so.1.24.3 -r--r--r-- 1 root root 136452 May 18 2012 libobjc.a lrwxrwxrwx 1 root root 31 May 18 2012 libobjc.so -> /usr/local/lib/libobjc.so.4.6.0 lrwxrwxrwx 1 root root 31 May 18 2012 libobjc.so.4 -> /usr/local/lib/libobjc.so.4.6.0 lrwxrwxrwx 1 root root 31 May 18 2012 libobjc.so.4.6 -> /usr/local/lib/libobjc.so.4.6.0 -r--r--r-- 1 root root 140820 May 18 2012 libobjc.so.4.6.0 lrwxrwxrwx 1 root root 33 May 18 2012 libobjcxx.so -> /usr/local/lib/libobjcxx.so.4.6.0 lrwxrwxrwx 1 root root 33 May 18 2012 libobjcxx.so.4 -> /usr/local/lib/libobjcxx.so.4.6.0 lrwxrwxrwx 1 root root 33 May 18 2012 libobjcxx.so.4.6 -> /usr/local/lib/libobjcxx.so.4.6.0 -r--r--r-- 1 root root 13002 May 18 2012 libobjcxx.so.4.6.0 drwxrwsr-x 4 root staff 4096 Mar 3 13:33 python2.7 address@hidden:/usr/local/lib$
I was looking yesterday night for a more recent tar ball, but couldn't locate one.
I'll recompile libobjc2 in debug and will update you. Thanks for the info David!
-- Laurent
David
On 5 Mar 2013, at 01:37, Laurent Michel <address@hidden> wrote:
PS/ I decided to recompile GNUstep simply because I was experiencing a crash with my compiled binary during the library initializations of GNUstep/libobjc. Namely, here is the backtrace:
[Thread debugging using libthread_db enabled] Using host libthread_db library "/lib/i386-linux-gnu/libthread_db.so.1". [New Thread 0xb7021b40 (LWP 8378)]
Program received signal SIGSEGV, Segmentation fault. 0xb7a58d50 in init_protocols () from /usr/local/lib/libobjc.so (gdb) where #0 0xb7a58d50 in init_protocols () from /usr/local/lib/libobjc.so #1 0xb7a58c47 in init_protocols () from /usr/local/lib/libobjc.so #2 0xb7a588c1 in objc_init_protocols () from /usr/local/lib/libobjc.so #3 0xb7a517e3 in objc_try_load_category () from /usr/local/lib/libobjc.so #4 0xb7a57251 in __objc_exec_class () from /usr/local/lib/libobjc.so #5 0xb76d15d4 in .objc_load_function () from /home/ldm/work/platform/Clo/queens2/../ORFoundation/libORFoundation.so #6 0xb7fece9b in ?? () from /lib/ld-linux.so.2 #7 0xb7fecf84 in ?? () from /lib/ld-linux.so.2 #8 0xb7fdf20f in ?? () from /lib/ld-linux.so.2 (gdb) quit A debugging session is active.
-- This email complies with ISO 3103
_______________________________________________ Gnustep-dev mailing list address@hidden https://lists.gnu.org/mailman/listinfo/gnustep-dev
_______________________________________________ Gnustep-dev mailing list address@hidden https://lists.gnu.org/mailman/listinfo/gnustep-dev
-- Sent from my Difference Engine
|