User tests: Successful: Unsuccessful:
If we haven't performed the UTF8MB4 conversion on core tables then we should still strip utf8mb4 characters
Update to Joomla 3.5 beta 5 through extension manager on a utf8mb4 compatible server (do not manually apply the database fixer in extension manager to do the utf8mb4 upgrade). Try and insert some 4 byte characters into the database in a sentence - at first this will cause data loss (all characters will be stripped after the 4 byte character will vanish) as the server hasn't been utf8mb4 provisioned yet. Apply patch and perform the same test -the utf8mb4 character will now be replaced with \xe2\xAf\x91
character (reference: http://www.utf8-chartable.de/unicode-utf8-table.pl?start=11136&number=128&names=-&utf8=string-literal) and you won't loose any other characters.
Do the database extension fix to upgrade the tables to utf8mb4 - the 4 byte charaters can now be inserted without any issues.
Note this adds effectively a extra DB query to every single page load - is it really required for this edge case?
Status | New | ⇒ | Pending |
Labels |
Added:
?
|
So parts 2 and 3 are good for me (amended test instructions as was my bad - we don't strip the utf8mb4 put replace it with that "box" symbol). I don't understand why you get the question mark in the first step tho :/
Don't know, this is the test server info:
Tests made in chrome.
I have tested this item successfully on dd90196
My test looked exactly the same as andre's, i.e. all as described, and in step 1 a "?".
Database Version: 5.5.47-0+deb7u1-log
Database Collation: utf8_general_ci
Database Connection Collation: utf8mb4_general_ci when looking after the conversion.
PHP Version: 5.6.18
Browser: Firefox 44.0.2
For me this is a positive test result, because the new code after patch works as described, even if the thing in general is worth a discussion.
I did not see any performance problems when testing, so maybe we could live with it like done with this PR?
And I would not rely on the conversion always be successful after using update component to update.
Also I would not say we should not stop support updating with extension installer anymore as long as we have it working (except of this necessary database fix to do the conversion). Some people still might need it, and so they will also need this PR.
Category | ⇒ | Libraries |
Status | Pending | ⇒ | Closed |
Closed_Date | 0000-00-00 00:00:00 | ⇒ | 2016-03-11 00:01:10 |
Closed_By | ⇒ | wilsonge |
Deleted previous comments and made new test. I don't see any characters disappearing in any case.
Done
Result:
Result
BTW used always the same 4 byte character????