* Ingo Strüwing <Ingo.Struewing@stripped> [10/02/03 21:06]:
> one backup test relies on the ability of dynamic arrays to do a memory
> allocation for every insert.
> In 6.0 this could be done by setting initial allocation to 0 and
> allocation increment to 1. This capability had been introduced by
Not just ok, but needed.
This is also useful for my_hash_init that would not allocate
anything (right now we have to use bzero instead of my_hash_init
because there is no way to init the hash without allocating
I would like to look at the patch. Who did it?
> In 5.6 it is still so that an initial allocation is always done. The
> first insert into a dynamic array does never cause a memory allocation
> (unless the initial attempt failed). This makes testing more complicated.
> Do you agree with backporting of the above mentioned patch? However, I
> would port only the changes to dynamic arrays. The changes to deadlock
> detection presume files, which would need to be added by porting one or
> more other patches.