shit, you hit the nail on the head. they were allocated from a Julia buffer, which was probabily GC'ed.
Thanks!!1 On Thursday, February 11, 2016 at 1:14:52 PM UTC-8, Yichao Yu wrote: > > On Thu, Feb 11, 2016 at 1:59 PM, Gabriel Goh <[email protected] > <javascript:>> wrote: > > Let's say I have a program which returns > > > > data = ccall(getcfunction()) > > > > data is a pointer to a C-struct, > > > > dataload = unsafe_load(data) > > > > which in turn contains a field a which is a 2D array, a pointer to some > > pointers of size 4, say. > > > > julia> pointer_to_array(dataload.a,4) > > Ptr{Node} @0x3b8f66c0 > > Ptr{Node} @0x3b8f66f0 > > Ptr{Node} @0x3b8f6750 > > Ptr{Node} @0x3b8f6810 > > > > Each pointer has an array of size 2. So this is a 4x2 2 dimensional > array > > > > julia> [pointer_to_array(L,2) for L in pointer_to_array(dataload.a,4)] > > [Node(1,9.424044618506457),Node(2,11.88905969931336)] > > [Node(1,9.678805056283334),Node(2,9.845676112951852)] > > [Node(1,8.210282978159183),Node(2,11.911456933936952)] > > [Node(1,11.916394646460297),Node(2,10.57806221488004)] > > > > My problem is that the data the second pointers are pointing to seems to > be > > invisible to the garbage collector, and gets corrupted once in a while. > > > > Turning off garbage collection seems to fix the problem, and forcing > more > > garbage collections increase the severity of it. > > > > Is there an easy way to "tell" julia that I want this whole struct > > untouched? > > I assume by "the second pointers" you mean the memory pointed to by > `L` in the code above. If that is the case, the question is how is the > memory they are pointing to being managed. Are they allocated by the C > function and needs to be manually free'd (and when are you freeing it > now) or are they pointing to a julia buffer you allocate? > > > > > Thanks! > > > > Gabe >
