David Ribeiro Alves has posted comments on this change. Change subject: KUDU-1931: fix SIGSEGV in file cache RandomAccessFile::memory_footprint() ......................................................................
Patch Set 1: (1 comment) http://gerrit.cloudera.org:8080/#/c/6359/1/src/kudu/util/file_cache.cc File src/kudu/util/file_cache.cc: PS1, Line 350: // Some anecdotal memory measurements taken inside gdb: : // - glibc 2.23 malloc_usable_size() on make_shared<FileType>: 88 bytes. : // - tcmalloc malloc_usable_size() on make_shared<FileType>: 96 bytes. : // - sizeof(std::shared_ptr::_Sp_counted_base): 16 bytes. : // - sizeof(*this): 72 bytes. > If you look at some other accounted objects, you'll see that they either ex What I meant was that there could easily be situations where a class is both with and without a shared_ptr, and if it doesn't inherit from RefCounted there's no way to account for the different uses. Particularly I was saying that that I'm ok with not accounting for the control block given the current uses of this method but that we should leave a note here (as an add on to the notes you already have) so that if we even add a memory_footprint() to FileChache we don't forget to add the shared_ptr control block size there as, at least in the fbm, that might become meaningful. -- To view, visit http://gerrit.cloudera.org:8080/6359 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-MessageType: comment Gerrit-Change-Id: I9e57c06dd35252d03729c10ff64fe43d79288513 Gerrit-PatchSet: 1 Gerrit-Project: kudu Gerrit-Branch: master Gerrit-Owner: Adar Dembo <[email protected]> Gerrit-Reviewer: Adar Dembo <[email protected]> Gerrit-Reviewer: Dan Burkert <[email protected]> Gerrit-Reviewer: David Ribeiro Alves <[email protected]> Gerrit-Reviewer: Kudu Jenkins Gerrit-Reviewer: Todd Lipcon <[email protected]> Gerrit-HasComments: Yes
