ygoldfeld

Follow

Yuri Goldfeld ygoldfeld

Follow

6 followers · 4 following

San Francisco Bay area

Achievements

Achievements

Popular repositories Loading

ygoldfeld doesn't have any public repositories yet.

474 contributions in the last year

Learn how we count contributions

Less

More

Contribution activity

April 2025

Created 12 commits in 3 repositories

Opened 4 pull requests in 3 repositories

Flow-IPC/ipc 2 merged

Pulling in submodules: <ipc_shm: Fix egregiously hidden and sneaky bug/typo in SHM-classic ipc::shm::classic::Pool_arena::construct(), which also potentially affects other consumers of SHM-classic, directly or otherwise: sometimes the memory taken from a reserved SHM-pool in SHM-classic was a few bytes fewer than actually necessary, leading to unpredictable behavior for certain object sizes. / Fixed small bug wherein SHM-classic ipc::shm::classic::Pool_arena logged as-if from component TRANSPORT instead of the correct component SHM. / In SHM-classic ipc::shm::classic::Pool_arena a couple internal changes for small memory/perf improvements: the boost.ipc index is unnecessary and is now omitted; and avoid a certain heap allocation. / Minor. (Opportunistic.)>
This contribution was made on Apr 10
(SHM-jemalloc) Removed an INFO-severity logged message that is logged from ipc::shm::arena_lend::jemalloc::Ipc_arena::get_collection_id(), a public API but more likely invoked internally from Shm_session::lend_object() which can be oft-used directly and/or for each capnp-encoded message sent through a SHM-jemalloc-backed ipc::transport::struc::Channel. / Use move-semantics in a few places internally to avoid unnecessary shared_ptr ref-count arithmetic. / Effectively remove a few redundant internal assert()s. / Comment and/or doc changes.
This contribution was made on Apr 3

Flow-IPC/ipc_shm 1 merged

Fix egregiously hidden and sneaky bug/typo in SHM-classic ipc::shm::classic::Pool_arena::construct(), which also potentially affects other consumers of SHM-classic, directly or otherwise: sometimes the memory taken from a reserved SHM-pool in SHM-classic was a few bytes fewer than actually necessary, leading to unpredictable behavior for certain object sizes. / Fixed small bug wherein SHM-classic ipc::shm::classic::Pool_arena logged as-if from component TRANSPORT instead of the correct component SHM. / In SHM-classic ipc::shm::classic::Pool_arena a couple internal changes for small memory/perf improvements: the boost.ipc index is unnecessary and is now omitted; and avoid a certain heap allocation. / Minor. (Opportunistic.)
This contribution was made on Apr 10

Flow-IPC/ipc_shm_arena_lend 1 merged

(SHM-jemalloc) Removed an INFO-severity logged message that is logged from ipc::shm::arena_lend::jemalloc::Ipc_arena::get_collection_id(), a public API but more likely invoked internally from Shm_session::lend_object() which can be oft-used directly and/or for each capnp-encoded message sent through a SHM-jemalloc-backed ipc::transport::struc::Channel. / Use move-semantics in a few places internally to avoid unnecessary shared_ptr ref-count arithmetic. / Effectively remove a few redundant internal assert()s. / Comment and/or doc changes.
This contribution was made on Apr 3

Created an issue in Flow-IPC/ipc that received 2 comments

ipc.session (+SHM-enabled): If using MQ transport in a SHM-enabled Session setup but with heap-based struc::Channel, msg size too small => barfs on moderately sized messages. / Deduced MQ msg size should be possinlenti override.

(CC @wkorbe) Per title: If using MQ transport in a SHM-enabled Session setup but with heap-based struc::Channel, msg size too small => barfs on mod…

2 comments

Opened 3 other issues in 2 repositories

Flow-IPC/ipc 2 open

(Future "breaking change" API improvement) Compile-time config for ipc::session objects to be done as Config object template parameters instead of long lists of individual per-option template parameters.
This contribution was made on Apr 3
Performance deep dive.
This contribution was made on Apr 3

Flow-IPC/ipc_shm 1 closed

(SHM-classic) Sneaky, evil typo bug! Arena::construct<T>() might silently allocate just a few bytes fewer than needed.
This contribution was made on Apr 10