GH-5059 remove calls to E(...) when the return value is already being checked #5066
Conversation
|
@kenwenzel I've removed all the calls to E(...) where the return value was already being checked. Take a look if you can and feel free to merge if you are happy with everything. You can also create a release if you want, if not then I might have time the week of the 8th of July. |
hmottestad
force-pushed
the
GH-5059-lmdb-reduce-calls-to-E
branch
from
ad3a1a3
to
935f515
Compare
|
Thank you. However, it helped us to find the empty key issue in the ValueStore. I will try to change that part to prevent looking up empty keys. |
There was a problem hiding this comment.
We could keep the checks for the put methods to detect any MDB_MAP_FULL errors.
There was a problem hiding this comment.
I don't have access to my computer for a good week. Feel free to modify the PR and merge when you're happy with it. Otherwise I'll look at it when I'm back.
| @@ -126,7 +126,7 @@ private synchronized boolean update(Object element, boolean add) throws IOExcept | |||
| keyVal.mv_data(keyBuf); | |||
|
|
|||
| if (add) { | |||
| if (E(mdb_put(factory.writeTxn, dbi, keyVal, dataVal, MDB_NOOVERWRITE)) == MDB_SUCCESS) { | |||
There was a problem hiding this comment.
revert this to some degree
There was a problem hiding this comment.
I added some simple retry logic to this one.
| @@ -873,14 +873,14 @@ public boolean storeTriple(long subj, long pred, long obj, long context, boolean | |||
| return recordCache.storeRecord(quad, explicit); | |||
| } | |||
|
|
|||
| int rc = E(mdb_put(writeTxn, mainIndex.getDB(explicit), keyVal, dataVal, MDB_NOOVERWRITE)); | |||
There was a problem hiding this comment.
revert this to some degree
There was a problem hiding this comment.
This is actually checked right below.
GitHub issue resolved: #5059
Briefly describe the changes proposed in this PR:
PR Author Checklist (see the contributor guidelines for more details):
mvn process-resourcesto format from the command line)