[MDEV-21288] innodb.full_crc32_import fails due to the use of optional compression algorithm Created: 2019-12-11 Updated: 2020-01-07 Resolved: 2020-01-07 |
|
| Status: | Closed |
| Project: | MariaDB Server |
| Component/s: | Tests |
| Affects Version/s: | 10.4 |
| Fix Version/s: | 10.4.12, 10.5.1 |
| Type: | Bug | Priority: | Major |
| Reporter: | Elena Stepanova | Assignee: | Sergei Golubchik |
| Resolution: | Fixed | Votes: | 0 |
| Labels: | None | ||
| Issue Links: |
|
||||
| Description |
|
The minimal fix would be
But while we are at it, I recommend
|
| Comments |
| Comment by Thirunarayanan Balathandayuthapani [ 2020-01-06 ] | ||||||||||||||||||||||
|
| ||||||||||||||||||||||
| Comment by Marko Mäkelä [ 2020-01-07 ] | ||||||||||||||||||||||
|
serg, can you please review this. I think that the build-time choice of allowed innodb_compression_algorithm constitutes a part of the file format. If our documented intention is that only innodb_compression_algorithm IN ('none','zlib') are guaranteed to be implemented and supported in all future versions, then the fix that was suggested by elenst is valid. However, if we intend to support more algorithms (2=lz4, 3=lzo, 4=lzma, 5=bzip2, 6=snappy), then the test should indeed keep failing and the build scripts should be fixed. My personal opinion is that it was a mistake to introduce so many options related to page_compressed. Especially bzip2 seems to be a bad match for InnoDB, because it is tuned for much larger input than the innodb_page_size. Hence, I would recommend to only promise future compatibility for innodb_compression_algorithm=zlib and to implement the suggested fix to the test case. | ||||||||||||||||||||||
| Comment by Marko Mäkelä [ 2020-01-07 ] | ||||||||||||||||||||||
|
I pushed the fix that was suggested by elenst. This relaxes the test, treating the compression algorithms as optional. To ensure file format compatibility, we might want to ensure that a wider selection of compressed page formats will be available at build time. |