Tom Ridge
Tom Ridge
Attached is a detailed log from running Zach's transfer-loop branch [pr1975.log](https://github.com/mirage/irmin/files/9093492/pr1975.log)
Zach reported that changing the gap_tolerance seemed to limit the number of transfer_loop iterations to a small number (2 or 3). It is not clear why gap_tolerance would affect the...
Within the transfer loop, Zach's code executes `Fm.reload`. This is likely to be slow. However, in order to pick up a possible file length change, with the current code, this...
These stats are significantly different from those reported in #1959.
I think it should be OK to just keep the LRU as it is (even with the GCed objects!); since the GCed objects are indexed by offset, they should no...
I think in the Tezos codebase they may do something to handle the case of (for example) RPCs attempting to access commits that have been GC'ed. So effectively with Tezos,...
Is this some idea to allow a close on a closed file to succeed without error? In my opinion, this is a bad decision. A close on an already closed...
Basically, allowing a close on an already closed file tends to indicate that the programmer isn't really sure about the resource lifetime. In general, the standard resource interfaces (such as:...
Looking at the implementation, and also at psq, I think I want a function like psq's "adjust".
I should add that in my application, all v have weight 1. So replacing k->v with k-> v' doesn't need to adjust weights or anything like that.