darrenldl / blockyarchive Goto Github PK

Improve code coverage of the following files via internal tests

• file_utils.rs
• integer_utils.rs
• misc_utils.rs
• multihash.rs
• rs_codec/encoder.rs
• rs_codec/repairer.rs
• sbx_block/header.rs
• sbx_block/metadata.rs
• sbx_block/mod.rs
• sbx_specs.rs
• time_utils.rs

Other files like *_core.rs, cli_*.rs, and functions that operate on files or require interaction with FS/OS are tested for code coverage by running debug binary through kcov with arguments attached
e.g. kcov --verify ... rsbx .... This is tracked by issue #58.

• calc
• check
• decode
• encode
• repair
• rescue
• show
• sort

• Add specs for burst error resistance level guessing
• Update specs for show core
• Add specs for repair core

Namely to avoid running out of disk space in travis vm

Update from 0.2 to 0.3

The file size calculatioin function fails to take burst gaps between metadata blocks into account.

For non RS enabled SBX versions, rsbx encode --no-meta ... will leave a blank spot in place of the metadata block.

Fixing this will require changing the SBX block write index calculation function.

Display after file size

Currently when RS is enabled, encode::encode_file keeps track of whether the current block set is full or not at the end of loop to decide whether to write data padding blocks.

This responsibility should be moved to rs_codec::encoder, with methods exposing number of empty slots, etc, to reduce work not directly related to encoding in encode_core.

Metadata blocks are counted twice, once for dummy write and one for final actual write

It mentions reading from log file, but show mode does not use log file, only rescue mode does.

rb on crates.io only supports single producer single consumer,
but ideally a multi producer and mutli consumer queue is used.

Say things like

"The configuration allows the container to survive up to {} burst SBX block errors, and each burst error may be up to {} blocks in size.", 1 + parity_shards, burst

Namely guessing procedure will only guess up to 1000.

If the level used during encoding is higher than 1000 then rsbx will make an incorrect guess.

Using the flag will cause the encode subcommand to display information then exit.

Information to display

• Input file name
• Output container name
• Output container version
• Output container block size
• Output container data size
• RS data count
• RS parity count
• Burst error resistance level
• Input file size
• Output container size
• Input file modification time

This should test all 6 versions.

Example

• ["test", "/"] should give "test" instead of "test/"

Fix:
Invoke RS codec at the end of loop in repair mode to try to repair the last set of missing blocks

E.g. rsbx calc 0 --no-meta should show metadata block count to be 0 instead of 1

Use either afl or honggfuzz

calc_required_len_and_seek_to_from_byte_range_inc calculation of seek_to may exceed last_possible_pos

Run rsbx through kcov for coverage of the following modes

• calc
• check
• decode
• encode
• repair
• rescue
• show
• sort

It should provide the following info

• Output container version (provided by user)
• Output container block size
• Output container data size
• Block distribution
  • Meta block count
  • Data block count
  • Parity block count
• RS related info
  • RS data count (provided by user)
  • RS parity count (provided by user)
  • Burst error resistance level (provided by user)
• Input file size (provided by user)
• Output container size