From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Date: Sun, 16 Jun 2019 17:57:00 +0300 From: Konstantin Osipov Subject: Re: [tarantool-patches] Re: [PATCH 09/10] key_def: pass alloc callback to key_def_dump_parts Message-ID: <20190616145700.GB26236@atlas> References: <20190518185231.GE9448@atlas> <20190601084142.GG29429@atlas> <20190610152821.cb6nb5vhcycx7nye@esperanza> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20190610152821.cb6nb5vhcycx7nye@esperanza> To: Vladimir Davydov Cc: tarantool-patches@freelists.org List-ID: * Vladimir Davydov [19/06/10 18:33]: > On Sat, Jun 01, 2019 at 11:41:42AM +0300, Konstantin Osipov wrote: > > * Konstantin Osipov [19/05/18 21:56]: > > > * Vladimir Davydov [19/05/17 17:54]: > > > > So that we can use an allocator different from the region in vylog. > > > > > > Please, let's simply always use malloc. > > > > > > btw, sql code is shit. not a single comment for entire call trace > > > of key_def_dump_parts() from there. I failed to understand whether > > > this is a hot path or not. > > > > I stand by this request. > > Please take a look at my reply I sent you earlier. Quoting it here: > > } I did it this way so as to use lsregion for allocating vylog records in > } the next patch. Are you suggesting to use malloc() instead? I guess > } that's possible and that'd simplify things quite a bit. I just wanted to > } use lock-free allocator for this. Not sure if it's really necessary > } though, because vy_log_tx_write isn't a hot path. It's totally OK to use malloc() for vy_log writes. -- Konstantin Osipov, Moscow, Russia