Stale cache when CTools is already enabled.

So, looks like the environemtn remains after acquia_search is uninstalled and re-installed? At the least, we should be able to work aroudn using db_merge or similar.

I don't see how this could be happening, unless the ctools code is saving the env a second time and somehow the drupal_static_reset() doesn't work properly?

This patch switches to db_merge and removes a bunch of duplicate or unneeded code.

the root cause is that the specific cache for ctools is not cleared when the load_all_environments is loaded. This can be avoided using :
ctools_export_load_object_reset('apachesolr_environment');

I did a bunch of testing and cache_clear_all was way too hard for this one, this simple snippet fixes the complete issue.

Also, I merged pwolanins suggestions and I added a reset flag to the load function so this cache can be cleared using the api and avoid further confusions for people that load the environment, save something and load it again and they do not see the change.

And just to clarify, I was able to replicate this issue and this patch fixes it for me. The changes from pwolanin are not necessary actually, as we could now, during save, say that it should not use the cache when loading the environment. However, I do like those changes so that is why I merged them

Seems like the reset should be handled at the top but just setting the static to NULL?

This might be better. Added all reset logic to the top and cleaned up the function a little bit. Not tested yet, doing so now

So the patch is not calling apachesolr_load_all_environments() with $reset any place?

#10: 1778050-10.patch queued for re-testing.

Basically its an API improvement. For example, I added it to clone now, but I can imagine more useful places?

The logic is wrong also if the tables are missing - you return NULL rather than array()

I think the logic here is better.

ok, I'm confused by Nick's changes and why they would make any difference to this bug. we never use the $reset flag, so ctools_export_load_object_reset() is never being called.

Seems like the missing piece is to clear the ctools cache in the save function too?

I started working on this patch since I wanted to know the root cause of the problem. it is true that the merge just overwrites the environment, but up to that moment we still did not understand why it would see the environment as $is_new while it was added just some milliseconds before.

The reason was this ctools cache, and this error only appears if you have ctools enabled. To get rid of this, I started adding a reset variable (before merging in your db_merge patch). I was using the reset with success, using the old $is_new variable.
When I merged your patch into this, this logic got removed.

Whenever someone does apachesolr_environment_save and then he loads that specific page a little later in the same page call, it could be that the page can't be found. To allow developers to avoid this problem, there is a $reset variable that allows to get a fresh load of all the environments, without ctools cache being in the way.

I would also be ok in adding this to the doxygen of that function as a caution/warning for developers. And only keep the reset in the save function.

Well, I'm happier to not add a $reset flag - if devs use the API functions correctly, it should never be neeeded.

here's such a simpler patch.

hurray for things that make life and code easier!

committed to 7.x

Need to decide how to handle the update or insert logic - this ought to be robust enough, even if not totally optimized.

missed a argument for the db_query()

This seems to work ok for me.

ugh - apparently I didn't actually commit #22!

committing now.