Hi Marti,
The subject was wrong in the previous emails. It should of course have been
"beta 1".
I've tried building the new beta 2, and 'make check' fails on 64-bit OSX: The
first call to DebugFree fails with "allocated by a different thread".
The reason is that the DebugMalloc/DebugFree functions assume that the
MemoryBlock struct doesn't have any padding between members. But on x86-64, the
struct is padded with 4 bytes after "KeepSize" as all pointers must be 8-byte
aligned (true for linux too).
typedef struct {
cmsUInt32Number KeepSize;
cmsContext WhoAllocated;
cmsUInt8Number mem[1];
} MemoryBlock;
So the calculation of the block header offset fails when using this define:
#define SIZE_OF_MEM_HEADER (sizeof(cmsUInt32Number) + sizeof(cmsContext))
Breaking in DebugFree, with (void*)Ptr argument = 0x100100160
(gdb) x/32x (0x100100160 - 16)
0x100100150: 0x00000038 0x00000000 0x00000001 0x00000000
0x100100160: 0x00000000 0x00000000 0x00000000 0x00000000
0x100100170: 0x00000000 0x00000000 0x00000000 0x00000000
Changing KeepSize to cmsUInt64Number (and updating the #define) makes the
testbed run successfully - and the returned memory will always be 8 and 16-byte
aligned. Only adding 12 bytes may break for 32-bit platforms that require
8-byte alignment for e.g. doubles, so always using 64 bit for the size might be
the easiest solution even for 32-bit platforms.
----
- Peter Speck
------------------------------------------------------------------------------
Download Intel® Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
_______________________________________________
Lcms-user mailing list
Lcms-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/lcms-user
