Mario Rugiero
Mario Rugiero
> I'm not against this but I personally like unix sockets better since you're sure no firewall bug can make the service reachable somewhere you don't expect, and FS permissions...
> This can be achieved by running blocky in its own network namespace, listening on localhost and then using `socat` or similar to proxy a unix socket to it. That's...
That sounds to me like something that should be handled by a different program and the client application. It's not only extra complexity on something that should be auditable, but...
> While I tend to be marginally careful about this, it is entirely possible we rely on more or less severe GNUisms across the codebase that would break in fun...
> I tried testing the PR, but that's not possible as we're in the middle of 1.5.0 upgrade. Will follow up once the upgrade is done. If possible, comment when...
Rebased and fixed the warnings.
Any benchmarks for comparison?
Answering myself from data in Slack: > Results for doubling an unsigned integer 10000000 times and assigning: > Time elapsed in double() is: 810.752416ms (OLD) > Time elapsed in double()...
@diegokingston I checked the failing pipeline. The tests pass fine, but codecov upload failed. I'm retrying to check if it's just a temporary service failure. Otherwise, I suspect it has...
I smell many more problems in the original code. `build_two_to_the_nth` does nothing close to what the comment says it does (it builds the equivalent to `0x111...11112` for a `FieldElement`, which...