281f06e01f
zstd streaming API was adding a null-block at end of frame for small input. Reason is : on small input, a single block is enough. ZSTD_CStream would size its input buffer to expect a single block of this size, automatically triggering a flush on reaching this size. Unfortunately, that last byte was generally received before the "end" directive (at least in `fileio`). The later "end" directive would force the creation of a 3-bytes last block to indicate end of frame. The solution is to not flush automatically, which is btw the expected behavior. It happens in this case because blocksize is defined with exactly the same size as input. Just adding one-byte is enough to stop triggering the automatic flush. I initially looked at another solution, solving the problem directly in the compression context. But it felt awkward. Now, the underlying compression API `ZSTD_compressContinue()` would take the decision the close a frame on reaching its expected end (`pledgedSrcSize`). This feels awkward, a responsability over-reach, beyond the definition of this API. ZSTD_compressContinue() is clearly documented as a guaranteed flush, with ZSTD_compressEnd() generating a guaranteed end. I faced similar issue when trying to port a similar mechanism at the higher streaming layer. Having ZSTD_CStream end a frame automatically on reaching `pledgedSrcSize` can surprise the caller, since it did not explicitly requested an end of frame. The only sensible action remaining after that is to end the frame with no additional input. This adds additional logic in the ZSTD_CStream state to check this condition. Plus some potential confusion on the meaning of ZSTD_endStream() with no additional input (ending confirmation ? new 0-size frame ?) In the end, just enlarging input buffer by 1 byte feels the least intrusive change. It's also a contract remaining inside the streaming layer, so the logic is contained in this part of the code. The patch also introduces a new test checking that size of small frame is as expected, without additional 3-bytes null block. |
||
|---|---|---|
| .. | ||
| files | ||
| fuzz | ||
| gzip | ||
| .gitignore | ||
| datagencli.c | ||
| decodecorpus.c | ||
| fullbench.c | ||
| fuzzer.c | ||
| invalidDictionaries.c | ||
| legacy.c | ||
| longmatch.c | ||
| Makefile | ||
| namespaceTest.c | ||
| paramgrill.c | ||
| playTests.sh | ||
| poolTests.c | ||
| README.md | ||
| roundTripCrash.c | ||
| seqgen.c | ||
| seqgen.h | ||
| symbols.c | ||
| test-zstd-speed.py | ||
| test-zstd-versions.py | ||
| zbufftest.c | ||
| zstreamtest.c | ||
Programs and scripts for automated testing of Zstandard
This directory contains the following programs and scripts:
datagen: Synthetic and parametrable data generator, for testsfullbench: Precisely measure speed for each zstd inner functionsfuzzer: Test tool, to check zstd integrity on target platformparamgrill: parameter tester for zstdtest-zstd-speed.py: script for testing zstd speed difference between commitstest-zstd-versions.py: compatibility test between zstd versions stored on Github (v0.1+)zbufftest: Test tool to check ZBUFF (a buffered streaming API) integrityzstreamtest: Fuzzer test tool for zstd streaming APIlegacy: Test tool to test decoding of legacy zstd framesdecodecorpus: Tool to generate valid Zstandard frames, for verifying decoder implementations
test-zstd-versions.py - script for testing zstd interoperability between versions
This script creates versionsTest directory to which zstd repository is cloned.
Then all tagged (released) versions of zstd are compiled.
In the following step interoperability between zstd versions is checked.
test-zstd-speed.py - script for testing zstd speed difference between commits
This script creates speedTest directory to which zstd repository is cloned.
Then it compiles all branches of zstd and performs a speed benchmark for a given list of files (the testFileNames parameter).
After sleepTime (an optional parameter, default 300 seconds) seconds the script checks repository for new commits.
If a new commit is found it is compiled and a speed benchmark for this commit is performed.
The results of the speed benchmark are compared to the previous results.
If compression or decompression speed for one of zstd levels is lower than lowerLimit (an optional parameter, default 0.98) the speed benchmark is restarted.
If second results are also lower than lowerLimit the warning e-mail is send to recipients from the list (the emails parameter).
Additional remarks:
- To be sure that speed results are accurate the script should be run on a "stable" target system with no other jobs running in parallel
- Using the script with virtual machines can lead to large variations of speed results
- The speed benchmark is not performed until computers' load average is lower than
maxLoadAvg(an optional parameter, default 0.75) - The script sends e-mails using
mutt; ifmuttis not available it sends e-mails without attachments usingmail; if both are not available it only prints a warning
The example usage with two test files, one e-mail address, and with an additional message:
./test-zstd-speed.py "silesia.tar calgary.tar" "email@gmail.com" --message "tested on my laptop" --sleepTime 60
To run the script in background please use:
nohup ./test-zstd-speed.py testFileNames emails &
The full list of parameters:
positional arguments:
testFileNames file names list for speed benchmark
emails list of e-mail addresses to send warnings
optional arguments:
-h, --help show this help message and exit
--message MESSAGE attach an additional message to e-mail
--lowerLimit LOWERLIMIT
send email if speed is lower than given limit
--maxLoadAvg MAXLOADAVG
maximum load average to start testing
--lastCLevel LASTCLEVEL
last compression level for testing
--sleepTime SLEEPTIME
frequency of repository checking in seconds
decodecorpus - tool to generate Zstandard frames for decoder testing
Command line tool to generate test .zst files.
This tool will generate .zst files with checksums, as well as optionally output the corresponding correct uncompressed data for extra verfication.
Example:
./decodecorpus -ptestfiles -otestfiles -n10000 -s5
will generate 10,000 sample .zst files using a seed of 5 in the testfiles directory,
with the zstd checksum field set,
as well as the 10,000 original files for more detailed comparison of decompression results.
./decodecorpus -t -T1mn
will choose a random seed, and for 1 minute, generate random test frames and ensure that the zstd library correctly decompresses them in both simple and streaming modes.