fix build with clang19 #895
Conversation
| @@ -375,7 +375,7 @@ BOOST_DATA_TEST_CASE(non_prune_to_prune, bdata::xrange(2) * bdata::xrange(2), en | |||
|
|
|||
| //upgrade to pruned... | |||
| t.conf = eosio::state_history::prune_config{ .prune_blocks = 4 }; | |||
| t.template check_n_bounce([]() {}); | |||
| t.check_n_bounce([]() {}); | |||
There was a problem hiding this comment.
For these, clang19 is more strict: these function calls need to be formatted as template foo<>(). But in all cases here we didn't need the template anyways.
| size_t tellp() { return this->pubseekoff(0, std::ios::cur); } | ||
| bool skip(size_t p) { this->pubseekoff(p, std::ios::cur); return true; } | ||
| size_t tellp() { return buf.pubseekoff(0, std::ios::cur); } | ||
| bool skip(size_t p) { buf.pubseekoff(p, std::ios::cur); return true; } |
There was a problem hiding this comment.
This is wild how it was never detected before.
There was a problem hiding this comment.
Probably because we never create a datastream with a class inheriting from std::streambuf.
Also, I believe that the way we use actually, on second thought, maybe it is fine.enable_if_t here is wrong, see the notes here.
There was a problem hiding this comment.
yeah I was a little worried the reason it was being flagged now was because something with the enable_if is goofed up and this was now being improperly selected.. but I couldn't find evidence of that actually occurring.
No description provided.