I haven't tracked down the details yet but I've been working on a site where on updating a node in some cases the node's cck fields are empty when saving a new revision. By implementing debugging in the hook_nodeapi call what we see is that on the case where the $op is update boost_nodeapi() is called but never returns. More debugging to do but filing this in case others are seeing it and can comment on what they've seen.
| Comment | File | Size | Author |
|---|---|---|---|
| #4 | boost-658942.patch | 997 bytes | mikeytown2 |
| #2 | 20091214-1mp12h8n2mmfy3inkb8kh232et.png | 168.81 KB | jbrauer |
Comments
Comment #1
mikeytown2 commentedhere's the code
I would have guessed 'insert' was causing the issue if you had flush views on node insert; since that loads up some views code and could have pushed PHP over the memory limit. Turn boost's verbose setting to 9 as that will allow you to trace the stack. Long story short this sounds like an out of memory error in PHP, try upping the memory limit, hopefully that takes care of this issue your having.
Comment #2
jbrauer commentedI think this may be a local environment interaction sort of thing though it should return an actionable error.
Attached is a section of the debugging log.
There is no PHP out of memory error thrown in this case. What I suspect is there may be a case where there is a database entry for the page but no file exists on disk to be purged. Still investigating that possibility and where it's dying. However, by commenting out the 'update' case of boost_nodeapi() the problem goes away (though obviously if there is a file present it is not removed from the boost cache.
Comment #3
mikeytown2 commenteddo you get anything written to the watchdog when you set boost verbose at 9; in regards to the update function? PHP shouldn't be error-ing out if Boost tries to delete a file that doesn't exist.
Comment #4
mikeytown2 commentedPut this patch in, and set verbose setting to level 9. Let me know what happens when you update a node.
Comment #5
mikeytown2 commented@jbrauer
is this bug a false report?
Comment #6
jbrauer commentedThis environment migrated away from using Boost so I don't have more information on this at this time. Marking at postponed needs more info but it would be reasonable to close/won't fix etc. it.
Comment #7
mikeytown2 commentedI believe this was fixed; had to do with this line at the top
If $node->nid doesn't exist then it would throw an error.