Merge pull request #1021 from AntelopeIO/GH-985-interrupt-apply

Do not unwind validated blocks on interrupt
AntelopeIO · Nov 13, 2024 · 88c8dbe · 88c8dbe
2 parents f871bf2 + c7b42ad
commit 88c8dbe
Show file tree

Hide file tree

Showing 2 changed files with 25 additions and 20 deletions.
diff --git a/libraries/chain/controller.cpp b/libraries/chain/controller.cpp
@@ -4438,8 +4438,9 @@ struct controller_impl {
             }
          }
 
-         auto start = fc::time_point::now();
+         const auto start_apply_blocks_loop = fc::time_point::now();
          for( auto ritr = new_head_branch.rbegin(); ritr != new_head_branch.rend(); ++ritr ) {
+            const auto start_apply_block = fc::time_point::now();
             auto except = std::exception_ptr{};
             const auto& bsp = *ritr;
             try {
@@ -4452,7 +4453,7 @@ struct controller_impl {
                   }
                   // Break every ~500ms to allow other tasks (e.g. get_info, SHiP) opportunity to run. User expected
                   // to call apply_blocks again if this returns incomplete.
-                  if (!replaying && fc::time_point::now() - start > fc::milliseconds(500)) {
+                  if (!replaying && fc::time_point::now() - start_apply_blocks_loop > fc::milliseconds(500)) {
                      result = controller::apply_blocks_result::incomplete;
                      break;
                   }
@@ -4463,7 +4464,11 @@ struct controller_impl {
                throw;
             } catch (const fc::exception& e) {
                if (e.code() == interrupt_exception::code_value) {
-                  ilog("interrupt while applying block ${bn} : ${id}", ("bn", bsp->block_num())("id", bsp->id()));
+                  if (fc::time_point::now() - start_apply_block < fc::milliseconds(2 * config::block_interval_ms)) {
+                     ilog("interrupt while applying block ${bn} : ${id}", ("bn", bsp->block_num())("id", bsp->id()));
+                     throw; // do not want to remove block from fork_db if not interrupting a long, maybe infinite, block
+                  }
+                  ilog("interrupt while applying block, removing block ${bn} : ${id}", ("bn", bsp->block_num())("id", bsp->id()));
                } else {
                   elog("exception thrown while applying block ${bn} : ${id}, previous ${p}, error: ${e}",
                        ("bn", bsp->block_num())("id", bsp->id())("p", bsp->previous())("e", e.to_detail_string()));
@@ -4480,18 +4485,20 @@ struct controller_impl {
                // Remove the block that threw and all forks built off it.
                fork_db.remove( (*ritr)->id() );
 
-               // pop all blocks from the bad fork, discarding their transactions
-               // ritr base is a forward itr to the last block successfully applied
-               auto applied_itr = ritr.base();
-               for( auto itr = applied_itr; itr != new_head_branch.end(); ++itr ) {
-                  pop_block();
-               }
-               EOS_ASSERT( !switch_fork || chain_head.id() == old_head_branch.back()->header.previous, fork_database_exception,
-                           "loss of sync between fork_db and chainbase during fork switch reversal" ); // _should_ never fail
+               if (switch_fork) {
+                  // pop all blocks from the bad fork, discarding their transactions
+                  // ritr base is a forward itr to the last block successfully applied
+                  auto applied_itr = ritr.base();
+                  for( auto itr = applied_itr; itr != new_head_branch.end(); ++itr ) {
+                     pop_block();
+                  }
+                  EOS_ASSERT( chain_head.id() == old_head_branch.back()->header.previous, fork_database_exception,
+                              "loss of sync between fork_db and chainbase during fork switch reversal" ); // _should_ never fail
 
-               // re-apply good blocks
-               for( auto ritr = old_head_branch.rbegin(); ritr != old_head_branch.rend(); ++ritr ) {
-                  apply_block( *ritr, controller::block_status::validated /* we previously validated these blocks*/, trx_lookup );
+                  // re-apply good blocks
+                  for( auto ritr = old_head_branch.rbegin(); ritr != old_head_branch.rend(); ++ritr ) {
+                     apply_block( *ritr, controller::block_status::validated /* we previously validated these blocks*/, trx_lookup );
+                  }
                }
                std::rethrow_exception(except);
             } // end if exception

diff --git a/tests/interrupt_trx_test.py b/tests/interrupt_trx_test.py
@@ -71,19 +71,17 @@
     trans=prodNode.pushMessage(contract, action, data, opts, silentErrors=True)
     assert trans and not trans[0], "push doitforever action did not fail as expected"
 
-    prodNode.processUrllibRequest("test_control", "swap_action", {"from": "doitslow", "to": "doitforever"})
-
-    action="doitslow"
-    trans=prodNode.pushMessage(contract, action, data, opts)
-    assert trans and trans[0], "Failed to push doitslow action"
-
     prodNode.waitForProducer("defproducera")
 
     prodNode.processUrllibRequest("test_control", "swap_action",
                                   {"from":"doitslow", "to":"doitforever",
                                    "trx_priv_key":EOSIO_ACCT_PRIVATE_DEFAULT_KEY,
                                    "blk_priv_key":cluster.defproduceraAccount.activePrivateKey})
 
+    action="doitslow"
+    trans=prodNode.pushMessage(contract, action, data, opts)
+    assert trans and trans[0], "Failed to push doitslow action"
+
     assert not prodNode.waitForHeadToAdvance(3), f"prodNode did advance head after doitforever action"
 
     prodNode.interruptAndVerifyExitStatus()