-
Notifications
You must be signed in to change notification settings - Fork 36k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
span: update constructors to match c++20 draft spec and add lifetimebound attribute #19387
Conversation
ACK apart from nit above. I've tested this by rebasing #13062 and #19326 on top, and compiling with Clang 10; no warnings. Adding an obvious It does seem that no warning is produced when |
Concept ACK: I think this will be useful also outside of diff --git a/src/attributes.h b/src/attributes.h
index 45099bd8b..9d5c1d44a 100644
--- a/src/attributes.h
+++ b/src/attributes.h
@@ -19,4 +19,14 @@
# endif
#endif
+#if defined(__clang__)
+# if __has_attribute(lifetimebound)
+# define LIFETIMEBOUND [[clang::lifetimebound]]
+# else
+# define LIFETIMEBOUND
+# endif
+#else
+# define LIFETIMEBOUND
+#endif
+
#endif // BITCOIN_ATTRIBUTES_H |
Concept ACK |
c++20's draft of std::span no longer includes move constructors.
f3f66a4
to
dafa901
Compare
Updated and squashed. @practicalswift I'd rather wait to add it there until there's a use for it, if that's ok. By the time we have it to use somewhere else, it may not be clang-only anymore.
Edit: whoops, |
See http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2018/p0936r0.pdf for reference. This helps to guard against dangling references caused by construction from temporaries such as: Span<const int> sp(std::vector<int>{1,2,3});
Sorry for the pushing/rebasing noise, done now.
@sipa It turns out that this warning does work for MakeSpan, but only in c++17 mode and above:
I've gone ahead and added the annotations there as well, so they should just magically work when we switch to c++17. |
I like magic. ACK 1d58cc7 |
Instead of magic, I'd rather remove them and add the corresponding constructors from c++20 when we switched to c++17 second-commit-only review ACK 1d58cc7 |
@@ -18,6 +18,16 @@ | |||
#define ASSERT_IF_DEBUG(x) | |||
#endif | |||
|
|||
#if defined(__clang__) | |||
#if __has_attribute(lifetimebound) | |||
#define SPAN_ATTR_LIFETIMEBOUND [[clang::lifetimebound]] |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
#define SPAN_ATTR_LIFETIMEBOUND [[clang::lifetimebound]] | |
#define ATTR_LIFETIMEBOUND [[clang::lifetimebound]] |
Can this be named a bit more generic, so that it can be used in other places such as #19426 ?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I namespaced it to avoid colliding with a more general define. As span.h
is designed to be an abstraction, I'd really prefer to keep it an island and not require that it include another header.
Since @sipa was quick to point out a few places where this could already be useful, how about going with @practicalswift's idea of adding a second, more generic define in attributes.h
?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@theuni Not sure how to weigh code duplication vs. a dependency on attributes.h. I'm fine with either.
@MarcoFalke Well, no reason why we can't have both. The attribute as argument to |
There are a lot more cases in the codebase where a data type stores a reference or pointer to another object provided to the constructor exist. For example:
Several of these accept const lvalue references, and store them, and are at risk already. Others take in a mutable lvalue reference or pointers, but would make sense to support rvalue/universal references as input (so that temporaries can be passed to it). In all those cases, having a So I think that means we should just put it in attributes.h. There is plenty of potential for it. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Non-expert code review ACK 1d58cc7
Looked at the linked doc and other docs/implementations I could find, abseil/span.h etc. This compiled fine including with it added to src/attributes.h
. Unfortunately I am currently stuck on Debian Clang 6 which is less helpful.
Upgraded to Clang 11 and now see the -Wdangling warning mentioned in the PR description.
|
Co-authored-by: practicalswift <practicalswift@users.noreply.github.com>
ACK e3e7446 |
Sorry, I misclicked. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
ACK e3e7446 change since last review is adding [[clang::lifetimebound]]
as LIFETIMEBOUND
to src/attributes.h as suggested in #19387 (comment).
…pec and add lifetimebound attribute bitcoin@62733fe bitcoin@1d58cc7 bitcoin@e3e7446
…pec and add lifetimebound attribute bitcoin@62733fe bitcoin@1d58cc7 bitcoin@e3e7446
partial merge bitcoin#13697, bitcoin#18591, bitcoin#19387, bitcoin#18468: update constructors to match c++20 draft spec
a1f3aed util: make EncodeBase64 consume Spans (furszy) c36754f Add MakeUCharSpan, to help constructing Span<[const] unsigned char> (Pieter Wuille) 1f18199 Make uint256 Span-convertible by adding ::data() (Pieter Wuille) 0067590 Add lifetimebound to attributes for general-purpose usage (Cory Fields) 26eb256 span: add lifetimebound attribute (Cory Fields) 696e91f span: (almost) match std::span's constructor behavior (Cory Fields) da207d4 doc: Mention Span in developer-notes.md (Pieter Wuille) 138053d doc: Document Span pitfalls (Pieter Wuille) e2216d7 Add sanity check asserts to span when -DDEBUG (Pieter Wuille) ebcc978 Add Span constructors for arrays and vectors (Pieter Wuille) 38f105b Make pointer-based Span construction safer (Pieter Wuille) 33d5297 Make Span size type unsigned (Pieter Wuille) 17118e1 Support conversion between Spans of compatible types (Pieter Wuille) f36042f Span front() and back() methods and SpanPopBack function backport (furszy) 1b2e7c8 Add more methods to Span class (Pieter Wuille) Pull request description: Another decouple from #2411. Purely focused on updating the Span sources to be up-to-date with current upstream. Following PRs/commits were back ported: * bitcoin#13697 (only 29943a9) * 2b0fcff (only span changes). * bitcoin#18591 (only 0fbde48). * bitcoin#18468 (without 2676aea). * bitcoin#19367. * bitcoin#19387. * bitcoin#19326 (only 5678250 and e63dcc3 here). * bitcoin#19687 (only e2aa1a5 here, 2bc2071 is inside #2411 and require net related commits that are introduced down the commits line there). Extra side note: Some of the current serialization compiler warnings in master will be fixed with this, other ones are coming down 2411 commits line as they need other previous PRs. ACKs for top commit: random-zebra: Tested ACK a1f3aed Tree-SHA512: 256bc2064724ea6d4f6fac722fafb4ed3e2b4590cdad61bf1e4a5be25e0632bafcff39235ea6ed7badbef4f79dbb4f866356372a1b3c201af23873884baedfea
ecde04a [Consensus] Bump Active Protocol version to 70923 for v5.3 (random-zebra) b63e4f5 Consensus: Add v5.3 enforcement height for testnet. (furszy) f44be94 Only relay IPv4, IPv6, Tor addresses (Pieter Wuille) 015298c fix: tor: Call event_base_loopbreak from the event's callback (furszy) 34ff7a8 Consensus: Add mnb ADDRv2 guard. (furszy) b4515dc GUI: Present v3 onion addresses properly in MNs list. (furszy) 337d43d tests: don't export in6addr_loopback (Vasil Dimov) 2cde8e0 GUI: Do not show the tor v3 onion address in the topbar. (furszy) 0b5f406 Doc: update tor.md with latest upstream information. (furszy) 89df7f2 addrman: ensure old versions don't parse peers.dat (Vasil Dimov) bb90c5c test: add getnetworkinfo network name regression tests (Jon Atack) d8e01b5 rpc: update GetNetworksInfo() to not return unsupported networks (Jon Atack) 57fc7b0 net: update GetNetworkName() with all enum Network cases (Jon Atack) 647d60b tests: Modify rpc_bind to conform to bitcoin#14532 behaviour. (Carl Dong) d4d6729 Allow running rpc_bind.py --nonloopback test without IPv6 (Kristaps Kaupe) 4a034d8 test: Add rpc_bind test to default-run tests (Wladimir J. van der Laan) 61a08af [tests] bind functional test nodes to 127.0.0.1 Prevents OSX firewall (Sjors Provoost) 6a4f1e0 test: Add basic addr relay test (furszy) 78aa61c net: Make addr relay mockable (furszy) ba954ca Send and require SENDADDRV2 before VERACK (Pieter Wuille) 61c2ed4 Bump net protocol version + don't send 'sendaddrv2' to pre-70923 software (furszy) ccd508a tor: make a TORv3 hidden service instead of TORv2 (Vasil Dimov) 6da9a14 net: advertise support for ADDRv2 via new message (furszy) e58d5d0 Migrate to test_large_inv() to Misbehaving logging. (furszy) d496b64 [QA] fix mininode CAddress ser/deser (Jonas Schnelli) cec9567 net: CAddress & CAddrMan: (un)serialize as ADDRv2 Change the serialization of `CAddrMan` to serialize its addresses in ADDRv2/BIP155 format by default. Introduce a new `CAddrMan` format version (3). (furszy) b8c1dda streams update: get rid of nType and nVersion. (furszy) 3eaa273 Support bypassing range check in ReadCompactSize (Pieter Wuille) a237ba4 net: recognize TORv3/I2P/CJDNS networks (Vasil Dimov) 8e50853 util: make EncodeBase32 consume Spans (Sebastian Falbesoner) 1f67e30 net: CNetAddr: add support to (un)serialize as ADDRv2 (Vasil Dimov) 2455420 test: move HasReason so it can be reused (furszy) d41adb4 util: move HasPrefix() so it can be reused (Vasil Dimov) f6f86af Unroll Keccak-f implementation (Pieter Wuille) 45222e6 Implement keccak-f[1600] and SHA3-256 (Pieter Wuille) 08ad06d net: change CNetAddr::ip to have flexible size (furszy) 3337219 net: improve encapsulation of CNetAddr. (furszy) 910d5c4 test: Do not instantiate CAddrDB for static call (Hennadii Stepanov) 6b607ef Drop IsLimited in favor of IsReachable (Ben Woosley) a40711b IsReachable is the inverse of IsLimited (DRY). Includes unit tests (marcaiaf) 8839828 net: don't accept non-left-contiguous netmasks (Vasil Dimov) 5d7f864 rpcbind: Warn about exposing RPC to untrusted networks (Luke Dashjr) 2a6abd8 CNetAddr: Add IsBindAny method to check for INADDR_ANY (Luke Dashjr) 4fdfa45 net: Always default rpcbind to localhost, never "all interfaces" (Luke Dashjr) 31064a8 net: Minor accumulated cleanups (furszy) 9f9c871 tests: Avoid using C-style NUL-terminated strings as arguments (practicalswift) f6c52a3 tests: Add tests to make sure lookup methods fail on std::string parameters with embedded NUL characters (practicalswift) a751b9b net: Avoid using C-style NUL-terminated strings as arguments in the netbase interface (furszy) f30869d test: add IsRFC2544 tests (Mark Tyneway) ed5abe1 Net: Proper CService deserialization + GetIn6Addr return false if addr isn't an IPv6 addr (furszy) 86d73fb net: save the network type explicitly in CNetAddr (Vasil Dimov) ad57dfc net: document `enum Network` (Vasil Dimov) cb160de netaddress: Update CNetAddr for ORCHIDv2 (Carl Dong) c3c04e4 net: Better misbehaving logging (furszy) 3660487 net: Use C++11 member initialization in protocol (Marco) 082baa3 refactor: Drop unused CBufferedFile::Seek() (Hennadii Stepanov) e2d776a util: CBufferedFile fixes (Larry Ruane) 6921f42 streams: backport OverrideStream class (furszy) Pull request description: Conjunction of a large number of back ports, updates and refactorings that made with the final goal of implementing v3 Onion addresses support (BIP155 https://github.com/bitcoin/bips/blob/master/bip-0155.mediawiki) before the tor v2 addresses EOL, scheduled, by the Tor project, for (1) July 15th: v2 addr support removal from the code base, and (2) October 15th: v2 addr network disable, where **every peer in our network running under Tor will loose the connection and drop the network**. As BIP155 describes, this is introducing a new P2P message to gossip longer node addresses over the P2P network. This is required to support new-generation Onion addresses, I2P, and potentially other networks that have longer endpoint addresses than fit in the 128 bits of the current addr message. In order to achieve the end goal, had to: 1. Create Span class and push it up to latest Bitcoin implementation. 2. Update the whole serialization framework and every object using it up to latest Bitcoin implementation (3-4 years ahead of what we currently are in master). 3. Update the address manager implementing ASN-based bucketing of the network nodes. 4. Update and refactor the netAddress and address manager tests to latest Bitcoin implementation (4 years ahead of what we currently are in master). 5. Several util string, vector, encodings, parsing, hashing backports and more.. Important note: This PR it is not meant to be merged as a standalone PR, will decouple smaller ones moving on. Adding on each sub-PR its own description isolated from this big monster. Second note: This is still a **work-in-progress**, not ready for testing yet. I'm probably missing to mention few PRs that have already adapted to our sources. Just making it public so can decouple the changes, we can start merging them and i can continue working a bit more confortable (rebase a +170 commits separate branch is not fun..). ### List of back ported and adapted PRs: Span and Serialization: ---------------- * bitcoin#12886. * bitcoin#12916. * bitcoin#13558. * bitcoin#13697. (Only Span's commit 29943a9) * bitcoin#17850. * bitcoin#17896. * bitcoin#12752. * bitcoin#16577. * bitcoin#16670. (without faebf62) * bitcoin#17957. * bitcoin#18021. * bitcoin#18087. * bitcoin#18112 (only from 353f376 that we don't support). * bitcoin#18167. * bitcoin#18317. * bitcoin#18591 (only Span's commit 0fbde48) * bitcoin#18468. * bitcoin#19020. * bitcoin#19032. * bitcoin#19367. * bitcoin#19387. Net, NetAddress and AddrMan: ---------------- * bitcoin#7932. * bitcoin#10756. * bitcoin#10765. * bitcoin#12218. * bitcoin#12855. * bitcoin#13532. * bitcoin#13575. * bitcoin#13815. * bitcoin#14532. * bitcoin#15051. * bitcoin#15138. * bitcoin#15689. * bitcoin#16702. * bitcoin#17243. * bitcoin#17345. * bitcoin#17754. * bitcoin#17758. * bitcoin#17812. * bitcoin#18023. * bitcoin#18454. * bitcoin#18512. * bitcoin#19314. * bitcoin#19687 Keys and Addresses encoding: ---------------- * bitcoin#11372. * bitcoin#17511. * bitcoin#17721. Util: ---------------- * bitcoin#9140. * bitcoin#16577. * bitcoin#16889. * bitcoin#19593. Bench: ---------------- * bitcoin#16299. BIP155: ---------------- * bitcoin#19351. * bitcoin#19360. * bitcoin#19534. * bitcoin#19628. * bitcoin#19841. * bitcoin#19845. * bitcoin#19954. * bitcoin#19991 (pending). * bitcoin#19845. * bitcoin#20000 (pending). * bitcoin#20120. * bitcoin#20284. * bitcoin#20564. * bitcoin#21157 (pending). * bitcoin#21564 (pending). * Fully removed v2 onion addr support. * Add hardcoded seeds. * Add release-notes, changes to files.md and every needed documentation. I'm currently working on the PRs marked as "pending", this isn't over, but I'm pretty pretty close :). What a long road.. ACKs for top commit: random-zebra: utACK ecde04a Fuzzbawls: ACK ecde04a Tree-SHA512: 82c95fbda76fce63f96d8a9af7fa9a89cb1e1b302b7891e27118a6103af0be23606bf202c7332fa61908205e6b6351764e2ec23d753f1e2484028f57c2e8b51a
…ound attribute Summary: > span: (almost) match std::span's constructor behavior > > c++20's draft of std::span no longer includes move constructors. > span: add lifetimebound attribute > > See http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2018/p0936r0.pdf for > reference. > > This helps to guard against dangling references caused by construction from > temporaries such as: > > `Span<const int> sp(std::vector<int>{1,2,3});` > Add lifetimebound to attributes for general-purpose usage > > Co-authored-by: practicalswift <practicalswift@users.noreply.github.com> This is a backport of [[bitcoin/bitcoin#19387 | core#19387]] Test Plan: With clang and gcc: `ninja all check-all` Reviewers: #bitcoin_abc, Fabien Reviewed By: #bitcoin_abc, Fabien Subscribers: Fabien Differential Revision: https://reviews.bitcoinabc.org/D10888
Replaces #19382 with a different approach. See this comment for the reasoning behind the switch.
--
Description from #19382:
See here for more detail on lifetimebound.
This is implemented using preprocesor macros rather than configure checks in order to keep span.h self-contained.
The
[[clang::lifetimebound]]
syntax was chosen over__attribute__((lifetimebound))
because the former is more flexible and works to guardthis
as well as function parameters, and also because at least for now, it's available only in clang.There are currently no violations in our codebase, but this can easily be tested by inserting one like this somewhere and compiling with a modern clang:
The result: