Re: Bug in base: NSMapGet()

[Fred Kiefer]

Hi Daniel,

could you please explain why you think this is a problem in base and not in 
libobjc2 or rather your code?

You are surely aware that the first call to printf("%p\n", NSMapGet(table, 
"xxx")); is actually the same as the later one with „yyy“?

I don’t use libobjc2 myself, so I cannot test your code here. From just looking 
at your code I see you use NSPointerFunctionsWeakMemory for the values. I would 
expect that this requires the values to be ObjC objects, but you add the C 
string „asdasd“. Shouldn’t you be using @„asdasd“ here?

This of course does not mean there could as well be a bug in 
NSConcreteMapTable, but you should try to provide better arguments to support 
this.

I just tried to run your code on with Xcode on my Mac and there it fails with 
the expected error message inside of NSMapInsert. And I really tried this after 
writing the arguments above.


Fred



> Am 03.06.2017 um 21:30 schrieb Daniel Ferreira (theiostream) <address@hidden>:
> 
> I believe I have found a bug trying out some basic behavior of
> NSMapGet(). When running the following program (compiled with clang
> and GNUstep's libobjc):
> 
> ---
> #import <Foundation/Foundation.h>
> 
> int main() {
>        NSPointerFunctionsOptions keyOptions =
> NSPointerFunctionsOpaqueMemory | NSPointerFunctionsOpaquePersonality;
>        NSPointerFunctionsOptions valueOptions =
> NSPointerFunctionsWeakMemory | NSPointerFunctionsObjectPersonality;
>        NSMapTable *table = [[NSMapTable alloc]
> initWithKeyOptions:keyOptions valueOptions:valueOptions capacity:0];
> 
>        printf("%p\n", NSMapGet(table, "xxx"));
>        NSMapInsert(table, "xxx", "asdasd");
>        printf("%p\n", NSMapGet(table, "xxx"));
>        printf("%p\n", NSMapGet(table, "yyy"));
> 
>        return 0;
> }
> ---
> 
> I get a segfault when trying to get a key that does not exist on the
> map table (in this case, 'yyy').
> 
> Stacktrace:
> 
> #0  objc_msgSend_fpret () at /root/gnustep/libobjc2/objc_msgSend.x86-64.S:130
> #1  0x00007ffff7bc5ba4 in autorelease (obj=0x4019df) at
> /root/gnustep/libobjc2/arc.m:277
> #2  0x00007ffff7bc5fa3 in objc_autorelease (obj=0x4019df) at
> /root/gnustep/libobjc2/arc.m:376
> #3  0x00007ffff7bc724d in objc_loadWeak (object=0x794b08) at
> /root/gnustep/libobjc2/arc.m:693
> #4  0x00007ffff75af773 in pointerFunctionsRead (PF=<optimized out>,
> addr=<optimized out>) at ././NSConcretePointerFunctions.h:105
> #5  GSIMapNodeForKeyInBucket (map=0x78cb20, bucket=0x794ac0, key=...)
> at ../Headers/GNUstepBase/GSIMap.h:639
> #6  0x00007ffff75acc65 in GSIMapNodeForKey (map=0x78cb20, key=...) at
> ../Headers/GNUstepBase/GSIMap.h:667
> #7  NSMapGet (table=0x78cb20, key=0x4019ea) at NSConcreteMapTable.m:591
> #8  0x00000000004009f2 in main ()
> 
> I tried investigating the issue and spent some hours trying to get the
> internals of NSConcreteMapTable, but after some hours I could come to
> no conclusion. Could anyone who played with this try to point me to a
> fix?
> 
> -- Daniel.
> 
> _______________________________________________
> Gnustep-dev mailing list
> address@hidden
> https://lists.gnu.org/mailman/listinfo/gnustep-dev