Method performUpdates

valdi/src/valdi/runtime/Resources/AssetsManager.cpp:900–938

}

void AssetsManager::retryFailedAssets() {
    auto guard = lock();

    // Capture before mutating _scheduledUpdates: if a pump is already pending, we only enqueue.
    const bool needScheduleUpdates = _pauseUpdatesCount == 0 && _scheduledUpdates.empty();

    bool anyReset = false;
    for (const auto& it : _assets) {
        const auto& managedAsset = it.second;
        if (managedAsset->getState() != AssetStateFailedRetryable) {
            continue;
        }

        managedAsset->setState(AssetStateInitial);

        // Consumers already received the failure and are marked notified, so re-resolving alone would
        // leave them untouched (getNextConsumerToUpdate skips notified consumers). Reset each failed
        // consumer to its initial, un-notified state so the re-resolve re-loads and re-notifies it,
        // exactly as a freshly added consumer would be handled.
        for (size_t i = 0; i < managedAsset->getConsumersSize(); i++) {
            const auto& consumer = managedAsset->getConsumer(i);
            if (consumer->getState() == AssetConsumerStateFailed) {
                consumer->setState(AssetConsumerStateInitial);
                consumer->setNotified(false);
            }
        }

        _scheduledUpdates.emplace_back(it.first);
        anyReset = true;
    }

    if (anyReset && needScheduleUpdates) {
        if (_mainThreadManager.currentThreadIsMainThread()) {
            performUpdates(std::move(guard));
        } else {
            guard.unlock();
            schedulePerformUpdates();
        }
    }
}

void AssetsManager::performUpdates(std::unique_lock<std::recursive_mutex>&& lock) {