Re: Memory leak on subquery as scalar operand

> On 2 May 2023, at 15:33, Daniel Gustafsson <daniel(at)yesql(dot)se> wrote:

> The attached 0005 is a WIP attempt to address a few of the FIXME's in this
> patchset on top of your the 0001-0004.

I've attached a rebased v5 which has the above fixes as well as a few smaller
ones missed in v4. Testing through the reports linked to from the open item
it's clear that while there is still a continuous rise in resident size, this
slows it down dramatically with the size levelling out which unpatched master
doesn't. Reducing LLVMJIT_LLVM_CONTEXT_REUSE_MAX back even further pushes the
levelling off further as well, making me wonder if this is something worth
using a GUC for given that we're stuck with the destructive LLVM linker for the
foreseeable future?

--
Daniel Gustafsson