From 24866b71c40f0fcb5a601d90d4f87366fe626090 Mon Sep 17 00:00:00 2001 From: John Ericson Date: Fri, 14 Apr 2023 18:18:32 -0400 Subject: [PATCH] Introduce `SingleDrvOutputs` In many cases we are dealing with a collection of realisations, they are all outputs of the same derivation. In that case, we don't need "derivation hashes modulos" to be part of our map key, because the output names alone will be unique. Those hashes are still part of the realisation proper, so we aren't loosing any information, we're just "normalizing our schema" by narrowing the "primary key". Besides making our data model a bit "tighter" this allows us to avoid a double `for` loop in `DerivationGoal::waiteeDone`. The inner `for` loop was previously just to select the output we cared about without knowing its hash. Now we can just select the output by name directly. Note that neither protocol is changed as part of this: we are still transferring `DrvOutputs` over the wire for `BuildResult`s. I would only consider revising this once #6223 is merged, and we can mention protocol versions inside factored-out serialization logic. Until then it is better not change anything because it would come a the cost of code reuse. --- src/build-remote/build-remote.cc | 5 ++-- src/libcmd/installables.cc | 4 +-- src/libstore/build-result.hh | 2 +- src/libstore/build/derivation-goal.cc | 26 ++++++++-------- src/libstore/build/derivation-goal.hh | 12 ++++---- src/libstore/build/goal.cc | 2 +- src/libstore/build/local-derivation-goal.cc | 6 ++-- src/libstore/build/local-derivation-goal.hh | 2 +- src/libstore/daemon.cc | 5 +++- src/libstore/legacy-ssh-store.cc | 6 +++- src/libstore/realisation.hh | 33 ++++++++++++++++++++- src/libstore/remote-store.cc | 20 +++++++++---- src/nix-store/nix-store.cc | 5 +++- 13 files changed, 90 insertions(+), 38 deletions(-) diff --git a/src/build-remote/build-remote.cc b/src/build-remote/build-remote.cc index cfc4baaca..ce9c7f45a 100644 --- a/src/build-remote/build-remote.cc +++ b/src/build-remote/build-remote.cc @@ -311,8 +311,9 @@ connected: auto thisOutputId = DrvOutput{ thisOutputHash, outputName }; if (!store->queryRealisation(thisOutputId)) { debug("missing output %s", outputName); - assert(result.builtOutputs.count(thisOutputId)); - auto newRealisation = result.builtOutputs.at(thisOutputId); + auto i = result.builtOutputs.find(outputName); + assert(i != result.builtOutputs.end()); + auto & newRealisation = i->second; missingRealisations.insert(newRealisation); missingPaths.insert(newRealisation.outPath); } diff --git a/src/libcmd/installables.cc b/src/libcmd/installables.cc index 32ae46d9f..0a2fe0073 100644 --- a/src/libcmd/installables.cc +++ b/src/libcmd/installables.cc @@ -593,8 +593,8 @@ std::vector, BuiltPathWithResult>> Installable::build std::visit(overloaded { [&](const DerivedPath::Built & bfd) { std::map outputs; - for (auto & path : buildResult.builtOutputs) - outputs.emplace(path.first.outputName, path.second.outPath); + for (auto & [outputName, realisation] : buildResult.builtOutputs) + outputs.emplace(outputName, realisation.outPath); res.push_back({aux.installable, { .path = BuiltPath::Built { bfd.drvPath, outputs }, .info = aux.info, diff --git a/src/libstore/build-result.hh b/src/libstore/build-result.hh index e07296eab..b7a56e791 100644 --- a/src/libstore/build-result.hh +++ b/src/libstore/build-result.hh @@ -87,7 +87,7 @@ struct BuildResult * For derivations, a mapping from the names of the wanted outputs * to actual paths. */ - DrvOutputs builtOutputs; + SingleDrvOutputs builtOutputs; /** * The start/stop times of the build (or one of the rounds, if it diff --git a/src/libstore/build/derivation-goal.cc b/src/libstore/build/derivation-goal.cc index 5bb664bff..a4bb94b0e 100644 --- a/src/libstore/build/derivation-goal.cc +++ b/src/libstore/build/derivation-goal.cc @@ -1013,7 +1013,7 @@ void DerivationGoal::resolvedFinished() auto resolvedDrv = *resolvedDrvGoal->drv; auto & resolvedResult = resolvedDrvGoal->buildResult; - DrvOutputs builtOutputs; + SingleDrvOutputs builtOutputs; if (resolvedResult.success()) { auto resolvedHashes = staticOutputHashes(worker.store, resolvedDrv); @@ -1039,7 +1039,7 @@ void DerivationGoal::resolvedFinished() worker.store.printStorePath(drvPath), wantedOutput); auto realisation = [&]{ - auto take1 = get(resolvedResult.builtOutputs, DrvOutput { *resolvedHash, wantedOutput }); + auto take1 = get(resolvedResult.builtOutputs, wantedOutput); if (take1) return *take1; /* The above `get` should work. But sateful tracking of @@ -1064,7 +1064,7 @@ void DerivationGoal::resolvedFinished() worker.store.registerDrvOutput(newRealisation); } outputPaths.insert(realisation.outPath); - builtOutputs.emplace(realisation.id, realisation); + builtOutputs.emplace(wantedOutput, realisation); } runPostBuildHook( @@ -1189,7 +1189,7 @@ HookReply DerivationGoal::tryBuildHook() } -DrvOutputs DerivationGoal::registerOutputs() +SingleDrvOutputs DerivationGoal::registerOutputs() { /* When using a build hook, the build hook can register the output as valid (by doing `nix-store --import'). If so we don't have @@ -1351,7 +1351,7 @@ OutputPathMap DerivationGoal::queryDerivationOutputMap() } -std::pair DerivationGoal::checkPathValidity() +std::pair DerivationGoal::checkPathValidity() { if (!drv->type().isPure()) return { false, {} }; @@ -1364,7 +1364,7 @@ std::pair DerivationGoal::checkPathValidity() return static_cast(names); }, }, wantedOutputs.raw()); - DrvOutputs validOutputs; + SingleDrvOutputs validOutputs; for (auto & i : queryPartialDerivationOutputMap()) { auto initialOutput = get(initialOutputs, i.first); @@ -1407,7 +1407,7 @@ std::pair DerivationGoal::checkPathValidity() } } if (info.wanted && info.known && info.known->isValid()) - validOutputs.emplace(drvOutput, Realisation { drvOutput, info.known->path }); + validOutputs.emplace(i.first, Realisation { drvOutput, info.known->path }); } // If we requested all the outputs, we are always fine. @@ -1431,7 +1431,7 @@ std::pair DerivationGoal::checkPathValidity() } -DrvOutputs DerivationGoal::assertPathValidity() +SingleDrvOutputs DerivationGoal::assertPathValidity() { auto [allValid, validOutputs] = checkPathValidity(); if (!allValid) @@ -1442,7 +1442,7 @@ DrvOutputs DerivationGoal::assertPathValidity() void DerivationGoal::done( BuildResult::Status status, - DrvOutputs builtOutputs, + SingleDrvOutputs builtOutputs, std::optional ex) { buildResult.status = status; @@ -1498,11 +1498,11 @@ void DerivationGoal::waiteeDone(GoalPtr waitee, ExitCode result) .outputs = OutputsSpec::Names { outputName }, }); if (buildResult.success()) { - for (auto & [output, realisation] : buildResult.builtOutputs) { + auto i = buildResult.builtOutputs.find(outputName); + if (i != buildResult.builtOutputs.end()) inputDrvOutputs.insert_or_assign( - { dg->drvPath, output.outputName }, - realisation.outPath); - } + { dg->drvPath, outputName }, + i->second.outPath); } } } diff --git a/src/libstore/build/derivation-goal.hh b/src/libstore/build/derivation-goal.hh index 9b5bd1805..7033b7a58 100644 --- a/src/libstore/build/derivation-goal.hh +++ b/src/libstore/build/derivation-goal.hh @@ -253,7 +253,7 @@ struct DerivationGoal : public Goal * Check that the derivation outputs all exist and register them * as valid. */ - virtual DrvOutputs registerOutputs(); + virtual SingleDrvOutputs registerOutputs(); /** * Open a log file and a pipe to it. @@ -306,17 +306,17 @@ struct DerivationGoal : public Goal * Update 'initialOutputs' to determine the current status of the * outputs of the derivation. Also returns a Boolean denoting * whether all outputs are valid and non-corrupt, and a - * 'DrvOutputs' structure containing the valid and wanted + * 'SingleDrvOutputs' structure containing the valid and wanted * outputs. */ - std::pair checkPathValidity(); + std::pair checkPathValidity(); /** * Aborts if any output is not valid or corrupt, and otherwise - * returns a 'DrvOutputs' structure containing the wanted + * returns a 'SingleDrvOutputs' structure containing the wanted * outputs. */ - DrvOutputs assertPathValidity(); + SingleDrvOutputs assertPathValidity(); /** * Forcibly kill the child process, if any. @@ -329,7 +329,7 @@ struct DerivationGoal : public Goal void done( BuildResult::Status status, - DrvOutputs builtOutputs = {}, + SingleDrvOutputs builtOutputs = {}, std::optional ex = {}); void waiteeDone(GoalPtr waitee, ExitCode result) override; diff --git a/src/libstore/build/goal.cc b/src/libstore/build/goal.cc index 13b2e509a..ca7097a68 100644 --- a/src/libstore/build/goal.cc +++ b/src/libstore/build/goal.cc @@ -23,7 +23,7 @@ BuildResult Goal::getBuildResult(const DerivedPath & req) { */ for (auto it = res.builtOutputs.begin(); it != res.builtOutputs.end();) { - if (bp.outputs.contains(it->first.outputName)) + if (bp.outputs.contains(it->first)) ++it; else it = res.builtOutputs.erase(it); diff --git a/src/libstore/build/local-derivation-goal.cc b/src/libstore/build/local-derivation-goal.cc index af937f6b1..6cb483a9c 100644 --- a/src/libstore/build/local-derivation-goal.cc +++ b/src/libstore/build/local-derivation-goal.cc @@ -2174,7 +2174,7 @@ void LocalDerivationGoal::runChild() } -DrvOutputs LocalDerivationGoal::registerOutputs() +SingleDrvOutputs LocalDerivationGoal::registerOutputs() { /* When using a build hook, the build hook can register the output as valid (by doing `nix-store --import'). If so we don't have @@ -2691,7 +2691,7 @@ DrvOutputs LocalDerivationGoal::registerOutputs() means it's safe to link the derivation to the output hash. We must do that for floating CA derivations, which otherwise couldn't be cached, but it's fine to do in all cases. */ - DrvOutputs builtOutputs; + SingleDrvOutputs builtOutputs; for (auto & [outputName, newInfo] : infos) { auto oldinfo = get(initialOutputs, outputName); @@ -2710,7 +2710,7 @@ DrvOutputs LocalDerivationGoal::registerOutputs() worker.store.registerDrvOutput(thisRealisation); } if (wantedOutputs.contains(outputName)) - builtOutputs.emplace(thisRealisation.id, thisRealisation); + builtOutputs.emplace(outputName, thisRealisation); } return builtOutputs; diff --git a/src/libstore/build/local-derivation-goal.hh b/src/libstore/build/local-derivation-goal.hh index 42d32a31a..9acd7593d 100644 --- a/src/libstore/build/local-derivation-goal.hh +++ b/src/libstore/build/local-derivation-goal.hh @@ -237,7 +237,7 @@ struct LocalDerivationGoal : public DerivationGoal * Check that the derivation outputs all exist and register them * as valid. */ - DrvOutputs registerOutputs() override; + SingleDrvOutputs registerOutputs() override; void signRealisation(Realisation &) override; diff --git a/src/libstore/daemon.cc b/src/libstore/daemon.cc index 63898f8dc..621a59c0a 100644 --- a/src/libstore/daemon.cc +++ b/src/libstore/daemon.cc @@ -637,7 +637,10 @@ static void performOp(TunnelLogger * logger, ref store, to << res.timesBuilt << res.isNonDeterministic << res.startTime << res.stopTime; } if (GET_PROTOCOL_MINOR(clientVersion) >= 28) { - worker_proto::write(*store, to, res.builtOutputs); + DrvOutputs builtOutputs; + for (auto & [output, realisation] : res.builtOutputs) + builtOutputs.insert_or_assign(realisation.id, realisation); + worker_proto::write(*store, to, builtOutputs); } break; } diff --git a/src/libstore/legacy-ssh-store.cc b/src/libstore/legacy-ssh-store.cc index 7b40b27e0..6e50fe6cd 100644 --- a/src/libstore/legacy-ssh-store.cc +++ b/src/libstore/legacy-ssh-store.cc @@ -294,7 +294,11 @@ public: if (GET_PROTOCOL_MINOR(conn->remoteVersion) >= 3) conn->from >> status.timesBuilt >> status.isNonDeterministic >> status.startTime >> status.stopTime; if (GET_PROTOCOL_MINOR(conn->remoteVersion) >= 6) { - status.builtOutputs = worker_proto::read(*this, conn->from, Phantom {}); + auto builtOutputs = worker_proto::read(*this, conn->from, Phantom {}); + for (auto && [output, realisation] : builtOutputs) + status.builtOutputs.insert_or_assign( + std::move(output.outputName), + std::move(realisation)); } return status; } diff --git a/src/libstore/realisation.hh b/src/libstore/realisation.hh index a18cf2aa8..3922d1267 100644 --- a/src/libstore/realisation.hh +++ b/src/libstore/realisation.hh @@ -13,9 +13,25 @@ namespace nix { class Store; +/** + * A general `Realisation` key. + * + * This is similar to a `DerivedPath::Opaque`, but the derivation is + * identified by its "hash modulo" instead of by its store path. + */ struct DrvOutput { - // The hash modulo of the derivation + /** + * The hash modulo of the derivation. + * + * Computed from the derivation itself for most types of + * derivations, but computed from the (fixed) content address of the + * output for fixed-output derivations. + */ Hash drvHash; + + /** + * The name of the output. + */ std::string outputName; std::string to_string() const; @@ -60,6 +76,21 @@ struct Realisation { GENERATE_CMP(Realisation, me->id, me->outPath); }; +/** + * Collection type for a single derivation's outputs' `Realisation`s. + * + * Since these are the outputs of a single derivation, we know the + * output names are unique so we can use them as the map key. + */ +typedef std::map SingleDrvOutputs; + +/** + * Collection type for multiple derivations' outputs' `Realisation`s. + * + * `DrvOutput` is used because in general the derivations are not all + * the same, so we need to identify firstly which derivation, and + * secondly which output of that derivation. + */ typedef std::map DrvOutputs; struct OpaquePath { diff --git a/src/libstore/remote-store.cc b/src/libstore/remote-store.cc index 734e6f27f..69e809a0f 100644 --- a/src/libstore/remote-store.cc +++ b/src/libstore/remote-store.cc @@ -152,7 +152,11 @@ BuildResult read(const Store & store, Source & from, Phantom _) >> res.isNonDeterministic >> res.startTime >> res.stopTime; - res.builtOutputs = worker_proto::read(store, from, Phantom {}); + auto builtOutputs = worker_proto::read(store, from, Phantom {}); + for (auto && [output, realisation] : builtOutputs) + res.builtOutputs.insert_or_assign( + std::move(output.outputName), + std::move(realisation)); return res; } @@ -165,7 +169,10 @@ void write(const Store & store, Sink & to, const BuildResult & res) << res.isNonDeterministic << res.startTime << res.stopTime; - worker_proto::write(store, to, res.builtOutputs); + DrvOutputs builtOutputs; + for (auto & [output, realisation] : res.builtOutputs) + builtOutputs.insert_or_assign(realisation.id, realisation); + worker_proto::write(store, to, builtOutputs); } @@ -941,10 +948,10 @@ std::vector RemoteStore::buildPathsWithResults( queryRealisation(outputId); if (!realisation) throw MissingRealisation(outputId); - res.builtOutputs.emplace(realisation->id, *realisation); + res.builtOutputs.emplace(output, *realisation); } else { res.builtOutputs.emplace( - outputId, + output, Realisation { .id = outputId, .outPath = outputPath, @@ -979,7 +986,10 @@ BuildResult RemoteStore::buildDerivation(const StorePath & drvPath, const BasicD } if (GET_PROTOCOL_MINOR(conn->daemonVersion) >= 28) { auto builtOutputs = worker_proto::read(*this, conn->from, Phantom {}); - res.builtOutputs = builtOutputs; + for (auto && [output, realisation] : builtOutputs) + res.builtOutputs.insert_or_assign( + std::move(output.outputName), + std::move(realisation)); } return res; } diff --git a/src/nix-store/nix-store.cc b/src/nix-store/nix-store.cc index 7035e6a7b..74f255bee 100644 --- a/src/nix-store/nix-store.cc +++ b/src/nix-store/nix-store.cc @@ -935,7 +935,10 @@ static void opServe(Strings opFlags, Strings opArgs) if (GET_PROTOCOL_MINOR(clientVersion) >= 3) out << status.timesBuilt << status.isNonDeterministic << status.startTime << status.stopTime; if (GET_PROTOCOL_MINOR(clientVersion) >= 6) { - worker_proto::write(*store, out, status.builtOutputs); + DrvOutputs builtOutputs; + for (auto & [output, realisation] : status.builtOutputs) + builtOutputs.insert_or_assign(realisation.id, realisation); + worker_proto::write(*store, out, builtOutputs); } break;