You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
util/benchharness.cc:211:17: error: comparison of integers of different signs: 'const unsigned long long' and 'const long long' [-Werror,-Wsign-compare]
if (nsecs < minNanoseconds) {
~~~~~ ^ ~~~~~~~~~~~~~~
1 error generated.
The text was updated successfully, but these errors were encountered:
… version (facebook#258)
* Revert "Add an option to disable write stall (facebook#251) (facebook#252)"
This reverts commit 32b29a1.
Signed-off-by: tabokie <xy.tao@outlook.com>
* use ignore_unknown_options regardless of RocksDB version
Signed-off-by: tabokie <xy.tao@outlook.com>
* disable only for lower version and fix test
Signed-off-by: tabokie <xy.tao@outlook.com>
luky116
pushed a commit
to luky116/rocksdb
that referenced
this issue
Jul 10, 2024
For current way of setting log number when applying manifest write, the log number could go backwards. Considering following sequence of events:
kMemtableSwitch with log number L
kMemtableSwitch with log number L+1
kManifestWrite with log number L (related to the first memtable switch)
If this is a follower and it takes over right after applying the manifest write, the alive log number will go backwards and be smaller than min_log_number_to_keep. This will cause all log numbers in alive_log_files_ to be pruned.
So we should only set the log number if it's strictly greater than existing alive log number. We don't need to do this for memtable switch though. Memtable switch should never set it to smaller value.
rocksdb 3.4 fails to build:
The text was updated successfully, but these errors were encountered: