Wouldn't it be a good idea to manually clone the leveldb subtree and use it for the compilation instead of always updating the leveldb/ directory in this repository? For better security, a commit hash from the other repo could be hard coded into the build system so the leveldb subtree repo can't be corrupted.
The downside would be that it would not be possible to compile offline but the DL process could be done at a very early stage like the configure
I fail to see which problem that would solve, ignoring the problems it introduces
Wouldn't it be a good idea to
You haven't explained why you think this would be better. Regardless, we aren't about to change our use of sub-modules.
Ok sorry forgot that.
The advantage would be that we don't need to always update the leveldb tree in this repository, we can directly use our fork.
How is updating the tree harder than updating the commit id of what the build system uses?
It's just an git subtree merge command.
There's no reason to change this, so let's leave it alone. We have 757 open issues and 303 PRs, please don't create problems where there are none.
(FWIW, the original reason to include leveldb in the tree, was that it's indirectly part of consensus. This is still the case)