I assume based on the comment this was something seen in the wild. However my links are not double serialized so I'm getting a bunch of these:
unserialize() expects parameter 1 to be string, array given CckLink.php:49
This is probably a bug somewhere in CCK which isn't really worth fixing since we can support both cases.
unserialize, check for array and unserialize again if its still a string.
Commit
N/A
N/A
N/A
| Comment | File | Size | Author |
|---|---|---|---|
| #15 | ccklinks_multiple-2588421-5.patch | 1.67 KB | thlor |
| #15 | mulitple_serialization_example.png | 40.96 KB | thlor |
| #10 | ccklinks_double-2588421-4.patch | 4.56 KB | neclimdul |
Comments
Comment #2
neclimdultest it bot.
Comment #3
phenaproximaThe committers probably won't accept this without a test.
Comment #4
neclimdulWhat about this? I didn't think ahead to setup making the interdiff but the test data is the only thing so the failure patch is basically an interdiff. Sorry.
Hopefully the double unserialize will throw the error and kill the test as expected.
Comment #6
neclimdulAwesome! Me getting the order of the patches wrong aside, that was as expected!
Comment #7
phenaproximaGreat. So the tests already catch this, but we didn't have any test data in our fixtures which would expose the problem.
Comment #10
neclimdulgugh... ok that didn't happen with pifr. reposting to work around the wonkyness.
Comment #12
webchickI looked for more information on why that double-escaping might happen, and traced it back to the initial commit: #2233883: Link migration needs to convert source url into the appropriate route format for storage Something in benjy's test data. Adam says he's seen it before too. So I was going to mark it needs work for a better explanation, but the explanation appears to be "gremlins" so... ;)
Since these changes are self-contained against an experimental module, I believe that makes this eligible for commit during RC. Tagging.
Committed and pushed to 8.0.x. Thanks!
Comment #13
neclimdulI really shouldn't gush
Webchick fixes things in a rush
Gremlins begone
She is too strong
She beats them with a git push.
Comment #15
thlor commentedWe are migrating a D6 site where we found that some attribute arrays are not only double serialized - one of them up to 19 times.
s:155:"s:146:"s:137:"s:128:"s:119:"s:110:"s:101:"s:93:"s:85:"s:77:"s:69:"s:61:"s:53:"s:45:"s:37:"s:29:"s:21:"s:13:"s:6:"a:0:{}";";";";";";";";";";";";";";";";";";";Now I didn't spend time to understand what causes this bug in Drupal 6, but I assume that whatever it is, it is related to whatever caused the bug reported originally in this issue. If you can advise how I could understand whether this is caused by a D6 core bug (making it necessary to tackle it in D8 migration), or caused by custom code in our D6 site, without having to spend too much time debugging EOL D6 code or sharing our D6 code base, please let me know :)
As a workaround, I refactored the last patch into a recursive function.
Comment #16
thlor commentedComment #18
anish.a commentedIs this issue still needs work?
Comment #19
thlor commented@anish.a yes. In the context of my project I solved it with the patch I submitted, but it would need to be implemented.
Comment #23
mikeryan@thlor The original issue here was committed to core a long time ago - a new issue should be opened to address any remaining issues in this area.
Comment #24
mikeryan