Re: [Tarantool-patches] [PATCH v2 0/3] Fiber storage leak

[Igor Munkin <imun@tarantool.org>]

Vlad,

Thanks, the patch is neat! I've looked on the new series since you added
one more patch to it. LGTM for the whole patchset.

On 16.01.20, Vladislav Shpilevoy wrote:
> The patchset makes fiber.storage be always deleted. Regardless of
> where was it created - in a fiber born in the Lua land, or in a
> fiber serving IProto requests and accessing Lua.
> 
> That removes
> 
>   - a leak occurred each time when fiber storage was created for
>     an IProto request;
> 
>   - a possibility to see previous request's artifacts left in
>     fiber.storage;
> 
> And makes possible to use fiber.storage as request-local data.
> 
> Branch: http://github.com/tarantool/tarantool/tree/gerold103/gh-4662-fiber-storage-leak
> Issue: https://github.com/tarantool/tarantool/issues/4662
> 
> Changes in v2:
> - Renamed back fiber.on_cleanup -> fiber.on_stop;
> - Removed dead code from box_process_call/eval(), which is now
>   handled by iproto.
> 
> Vladislav Shpilevoy (3):
>   fiber: unref fiber.storage via global Lua state
>   fiber: destroy fiber.storage created by iproto
>   box: remove dead code from box_process_call/eval()
> 
>  src/box/call.c                               | 28 +------
>  src/box/iproto.cc                            | 44 ++++++++--
>  src/box/txn.c                                |  1 +
>  src/lib/core/fiber.c                         | 18 +++-
>  src/lib/core/fiber.h                         | 14 +++-
>  src/lua/fiber.c                              | 35 ++++++--
>  test/app/gh-4662-fiber-storage-leak.result   | 88 ++++++++++++++++++++
>  test/app/gh-4662-fiber-storage-leak.test.lua | 43 ++++++++++
>  8 files changed, 230 insertions(+), 41 deletions(-)
>  create mode 100644 test/app/gh-4662-fiber-storage-leak.result
>  create mode 100644 test/app/gh-4662-fiber-storage-leak.test.lua
> 
> -- 
> 2.21.1 (Apple Git-122.3)
> 

-- 
Best regards,
IM