best effort writer never releases cache changes

[liamstask]

publisher config:

PublisherAttributes pub_attr;
    pub_attr.topic.topicKind = NO_KEY;
    pub_attr.topic.topicName = TopicData::topic_name;
    pub_attr.topic.topicDataType = TopicData::topic_type;
    pub_attr.topic.historyQos.kind = KEEP_LAST_HISTORY_QOS;
    pub_attr.topic.historyQos.depth = 1;
    pub_attr.topic.resourceLimitsQos.max_samples = 1;
    pub_attr.topic.resourceLimitsQos.allocated_samples = 50;
    pub_attr.times.heartbeatPeriod.seconds = 0;
    pub_attr.times.heartbeatPeriod.fraction = 4294967 * 100; // ~100 millis
    pub_attr.qos.m_reliability.kind = BEST_EFFORT_RELIABILITY_QOS;

subscriber config:

SubscriberAttributes sub_attr;
    sub_attr.topic.topicKind = NO_KEY;
    sub_attr.topic.topicName = TopicData::topic_name;
    sub_attr.topic.topicDataType = TopicData::topic_type;
    sub_attr.topic.historyQos.kind = KEEP_LAST_HISTORY_QOS;
    sub_attr.topic.historyQos.depth = 1;
    sub_attr.times.heartbeatResponseDelay.seconds = 0;
    sub_attr.times.heartbeatResponseDelay.fraction = 4294967 * 50; // ~50 millis;
    sub_attr.topic.resourceLimitsQos.max_samples = 1;
    sub_attr.topic.resourceLimitsQos.allocated_samples = 5;
    sub_attr.qos.m_reliability.kind = BEST_EFFORT_RELIABILITY_QOS;

After 50 samples (i.e. allocated_samples for the publisher) are written through the publisher, all subsequent writes fail and I see the following messages logged:

[Warning][CacheChangePool::allocateGroup]           Maximum number of allowed reserved caches reached
[Warning][RTPSWriter::new_change]                   Problem reserving Cache from the History

I never see CacheChangePool::release_Cache() called for this writer.

I have a local change to StatelessWriter::clean_history() to return true (per #34) to allow the writer to even attempt writing in the first place.

[liamstask]

One fix for this is to edit History::remove_all_changes() as described in #40 and call it from StatelessWriter::clean_history().

[SantiagoMunoz]

Hello @liamstask
Thanks pointing that out! I've been looking into the issue but it seems that fixing that mutating vector you discovered plus some of the work we have been putting in this month has solved it. In any case, I've also added a test to make sure and we will keep an eye on this just in case.

[liamstask]

@SantiagoMunoz - great! Has this fix already been pushed?

[SantiagoMunoz]

@liamstask We are aiming to release a new version in the next few days, I will keep you updated :)

[SantiagoMunoz]

New version is available on master. Cheers!