Fix use of rtxn without a mdb_txn_safe wrapper #8379
Conversation
|
this is not an ideal fix, just a quick'n'dirty to make sure we're on the right track. the problem with this patch is that it means TXN_PREFIX_RDONLY uses mdb_txn_safe redundantly. |
hyc
force-pushed
the
rtxnguard
branch
2 times, most recently
from
211a0e7 to
897ce7d
Compare
|
ok. I see that it's looping, will check it out. |
src/blockchain_db/blockchain_db.h
Outdated
| if (readonly) | ||
| db->block_rtxn_stop(); | ||
| else | ||
| if (!readonly) |
There was a problem hiding this comment.
Should this be if (!readonly && active) ?
There was a problem hiding this comment.
This is the only relevant comment, and this code is shippable otherwise.
There was a problem hiding this comment.
It is already only invoked if (active), see line 1886.
There was a problem hiding this comment.
I guess, to eliminate all doubt, could move the if (active) check here from there.
| #define TXN_PREFIX_RDONLY() \ | ||
| MDB_txn *m_txn; \ | ||
| mdb_txn_cursors *m_cursors; \ | ||
| mdb_txn_safe auto_txn; \ |
There was a problem hiding this comment.
Is this still needed if block_rtxn_start uses it too now? This should be why the code worked in most situations - the bulk of the calls were using TXN_PREFIX_RDONLY.
I don't think it hurts to double up, but it looks unnecessary.
There was a problem hiding this comment.
Oooops, looks like you commented on this above already. I figured it would work, but I think you'd have to go back through and do a major re-haul to "get it correct", whereas this needs push sooner rather than later.
Fix use of rtxn without a mdb_txn_safe wrapper