It has observed that, entitycache creates tables during install under hook_schema() but tables are not droped during hook_uninstall()
This causes issues when module is disabled and someone try to enable it again.
This may leads to erro like:
exception 'DatabaseSchemaObjectExistsException' with message 'Table cache_entity_file already exists.'
| Comment | File | Size | Author |
|---|---|---|---|
| #10 | 2897150-10.patch | 673 bytes | mcdruid |
| #7 | exception-cache_entity_-already-exist-2897150-7.patch | 891 bytes | greggles |
| #3 | exception-cache_entity_-already-exist-2897150-3.patch | 890 bytes | dineshw |
| #2 | exception-cache_entity_$type-already-exist-2897150-2.patch | 1.04 KB | dineshw |
Comments
Comment #2
dineshw commentedTrying to drop existing tables in case hook_schema() gets executed again.
Better approach could be using hook_uninstall().
Comment #3
dineshw commentedUpdated Patch
Comment #4
dineshw commentedComment #6
gregglesA small code style nitpick:
There needs to be a space after the `if` and before the `(`.
Comment #7
gregglesReroll to fix that.
Comment #9
mcdruid commentedHmm, as mentioned in #2 I'm not certain we want to be dropping all the existing tables every time
hook_schema()is invoked.Manually testing this patch and installing, enabling and disabling entitycache I'm still hitting errors when
drupal_flush_all_caches()tries to flush tables that no longer exist.We can keep adding workarounds (e.g. check if the table exists before adding each bin in
entitycache_flush_caches()) but I think perhaps it'd be better to try a slightly different approach to solving the problem.That may be as simple as dropping the tables in
hook_uninstall(), but we'll want to check if that covers enabling/disabling too.Comment #10
mcdruid commentedHere's a patch that drops the tables in
hook_uninstall().However, this doesn't actually seem to be necessary AFAICS.
https://api.drupal.org/api/drupal/modules%21system%21system.api.php/func... says:
...and I've verified that works okay with some manual testing (using drush).
When the module is disabled, the tables are not removed (which I think is the correct behaviour).
I don't see any errors when the module is enabled again.
So I'm not actually able to reproduce there being any problem with letting the install/enable/disable/uninstall API work as it does now alongside the existing
hook_schema()implementation.Can anyone provide any steps to reproduce an error that requires a change in entitycache?
Comment #11
mcdruid commented