Failure test case #750
Failure test case #750
Conversation
mvines
added
work in progress
|
nice job, I don't understand it yet. |
|
|
||
| // NOTE: if Entry grows to larger than a transaction, the code below falls over | ||
| let threshold = (BLOB_DATA_SIZE / 256) - 1; // 256 is transaction size | ||
| let threshold = (BLOB_DATA_SIZE / PACKET_DATA_SIZE) - 1; |
There was a problem hiding this comment.
PACKET_DATA_SIZE may be the wrong magic number, what you want is sizeof(Transaction)
the test also assumes that the rest of an Entry struct is smaller than a Transaction (hence the -1)
threshold should be (BLOB_DATA_SIZE - serialized_size(Entry)) / serialized_size(Transaction), assuming there's zero overhead for typing in serialization of an Entry (which holds a Vector of Transactions)
There was a problem hiding this comment.
Ideally we would use sizeof(serialized Transaction) no? But we don’t have a way to generate that other than serializing all the different types of Transactions at runtime. sizeof(Transaction), the in-memory representation, is more than 300 bytes. But when serialized always must be less than 256, which I think is what this const represents
|
I understand it now, working on a fix. |
|
❤️. I was going to write a fix after 🚲 but thanks! |
|
my inclination is to just do the /256 thing instead of messing around with will_fit() |
|
Subsumed by #751 |
Test that reproduces the crash at #740