diff options
author | Abseil Team <absl-team@google.com> | 2019-04-01T21·06-0700 |
---|---|---|
committer | Gennadiy Rozental <rogeeff@google.com> | 2019-04-01T22·15-0400 |
commit | 93dfcf74cb5fccae3da07897d8613ae6cab958a0 (patch) | |
tree | 04896cfc845d55a724a1aac5589caf2045fc33a9 /absl/container/internal/raw_hash_set.h | |
parent | 2c8421e1c6cef0da9e8a20b01c15256ec9ec116d (diff) |
Export of internal Abseil changes.
-- 855576faf9556573fd74c2874b290d8feb6565d5 by Gennadiy Rozental <rogeeff@google.com>: Import of CCTZ from GitHub. PiperOrigin-RevId: 241395451 -- b93bfd43eb2a992258f131e10f503526cfec6d48 by CJ Johnson <johnsoncj@google.com>: Fixes comment over AbslHashValue for InlinedVector PiperOrigin-RevId: 241368320 -- 75f58dafcac7d78c28d92a61ec7e53c5b3b86697 by Matt Kulukundis <kfm@google.com>: Do not call sampling logic for tables with custom allocators. PiperOrigin-RevId: 241356451 -- 09f1b4889476ff707a54189aff540e2fe1edcf61 by Derek Mauro <dmauro@google.com>: Re-enable optionalTest.InPlaceTSFINAEBug after libc++ update PiperOrigin-RevId: 241222673 -- 01a8bb5a8cb1e13e88ddb92f9c0160beb6e126be by Derek Mauro <dmauro@google.com>: Update Clang on Kokoro to r356196. This includes a workaround for a -Wgnu-include-next warning fixed by https://reviews.llvm.org/rG0706e144d57305782988dd4367530ae04986116f PiperOrigin-RevId: 241222395 -- 1de66bb669a7ec1494d6064677687f761ee2d369 by Abseil Team <absl-team@google.com>: Remove identical test and fix char to string per comment PiperOrigin-RevId: 240855512 GitOrigin-RevId: 855576faf9556573fd74c2874b290d8feb6565d5 Change-Id: Ie155b209ef5567e6597da6ef1844db7e2ad72586
Diffstat (limited to 'absl/container/internal/raw_hash_set.h')
-rw-r--r-- | absl/container/internal/raw_hash_set.h | 13 |
1 files changed, 12 insertions, 1 deletions
diff --git a/absl/container/internal/raw_hash_set.h b/absl/container/internal/raw_hash_set.h index 85e33344c7d5..9c926812a07a 100644 --- a/absl/container/internal/raw_hash_set.h +++ b/absl/container/internal/raw_hash_set.h @@ -1437,7 +1437,18 @@ class raw_hash_set { void initialize_slots() { assert(capacity_); - if (slots_ == nullptr) { + // Folks with custom allocators often make unwaranted assumptions about the + // behavior of their classes vis-a-vis trivial destructability and what + // calls they will or wont make. Avoid sampling for people with custom + // allocators to get us out of this mess. This is not a hard guarntee but a + // workaround while we plan the exact guarantee we want to provide. + // + // People are often sloppy with the exact type of their allocator (sometimes + // it has an extra const or is missing the pair, but rebinds made it work + // anyway). To avoid the ambiguitity, we work off SlotAlloc which we have + // bound more carefully. + if (std::is_same<SlotAlloc, std::allocator<slot_type>>::value && + slots_ == nullptr) { infoz_ = Sample(); } |