[#69084] [Ruby trunk - Feature #11124] [Open] [PATCH] lib/*: use monotonic clock for timeouts — normalperson@...
Issue #11124 has been reported by Eric Wong.
5 messages
2015/05/06
[#69138] [Ruby trunk - Feature #11136] [PATCH] webrick: avoid fcntl module — nobu@...
Issue #11136 has been updated by Nobuyoshi Nakada.
3 messages
2015/05/12
[#69160] [Ruby trunk - Feature #11146] [PATCH] variable.c: initialize generic_iv_tbl at start — nobu@...
Issue #11146 has been updated by Nobuyoshi Nakada.
4 messages
2015/05/13
[#69175] Re: [Ruby trunk - Feature #11146] [PATCH] variable.c: initialize generic_iv_tbl at start
— Eric Wong <normalperson@...>
2015/05/13
nobu@ruby-lang.org wrote:
[ruby-core:69082] [Ruby trunk - Bug #11123] [Open] [PATCH] bsock_recmvsg_internal needs GC guard for control buffer
From:
normalperson@...
Date:
2015-05-06 21:31:39 UTC
List:
ruby-core #69082
Issue #11123 has been reported by Eric Wong. ---------------------------------------- Bug #11123: [PATCH] bsock_recmvsg_internal needs GC guard for control buffer https://bugs.ruby-lang.org/issues/11123 * Author: Eric Wong * Status: Open * Priority: Normal * Assignee: Akira Tanaka * ruby -v: trunk * Backport: 2.0.0: REQUIRED, 2.1: REQUIRED, 2.2: REQUIRED ---------------------------------------- The control buffer may be used throughout the function, so prevent the string from being lost to GC. I'm pretty sure this is correct and we'll need it, I'll commit in a day or two unless I missed something. I was planning to introduce new features before I noticed this bug: * "exception: false" * explicit buffer reuse to reduce GC, similar to read_nonblock(num, str) I also feel 4K buffers are too big for stack and should use heap instead (as ALLOCV caps stack allocations at 1K) ---Files-------------------------------- 0001-ext-socket-ancdata.c-bsock_recvmsg_internal-GC-guard.patch (845 Bytes) 0001-ext-socket-ancdata.c-bsock_recvmsg_internal-GC-guard.patch (845 Bytes) -- https://bugs.ruby-lang.org/