Fix for failure of unsigned_joins test under centos8 #1822
Conversation
Debug logs Debug logs - fix typo Debug logs Debug logs debug logs debug logs debug logs debug logs debug logs debug logs debug logs debug logs Solution to test Remove debug logs
| throw (runtime_error("Bad column width in CP entry string")); | ||
| } | ||
| pTok = strtok(NULL, " "); | ||
|
|
There was a problem hiding this comment.
We need to add support for 128-bit min/max values as well here. The sending side BRMReporter::sendCPToFile either streams CPInfoMerge::max/CPInfoMerge::min or CPInfoMerge::bigMax/CPInfoMerge::bigMin based on CPInfoMerge::type and CPInfoMerge::colWidth. So the parser here should first parse out type and colWidth and then based on these 2 values, should either set the 128-bit or the 64-bit min/max values.
There was a problem hiding this comment.
Thanks!
It is not possible to have tests for that functionality using joins. I've created task specifically to address your comment: https://jira.mariadb.org/browse/MCOL-4652
|
Looks like the PR is already merged before my review. I have requested changes in the comment above. Commit message is also not correct. I would also add the MCOL number as the first thing in the PR title. |
The problem was that column width was not set during BRM reporting stage in cpimport processing. That made column width a garbage, prevented correct ranges to be set and as a result some extents were incorrectly skipped during SELECT operation.
The fixes are in we_brmupdater.cpp: a change to display proper "CP" line format in the comments in the BRM report file, a change to add print of column width into file and a change to parse column width.