Am 30.03.2019 um 16:07 hat Eric Blake geschrieben: > Disk sizes close to INT64_MAX cause overflow, for some pretty > ridiculous output: > > $ ./nbdkit -U - memory size=$((2**63 - 512)) --run 'qemu-img info $nbd' > image: nbd+unix://?socket=/tmp/nbdkitHSAzNz/socket > file format: raw > virtual size: -8388607T (9223372036854775296 bytes) > disk size: unavailable > > But there's no reason to have two separate implementations of integer > to human-readable abbreviation, where one has overflow and stops at > 'T', while the other avoids overflow and goes all the way to 'E'. With > this patch, the output now claims 8EiB instead of -8388607T, which > really is the correct rounding of largest file size supported by qemu > (we could go 511 bytes larger if we used byte-accurate sizing instead > of rounding up to the next sector boundary, but that wouldn't change > the human-readable result). > > Reported-by: Richard W.M. Jones <rjo...@redhat.com> > Signed-off-by: Eric Blake <ebl...@redhat.com>
This is quite obviously a bug fix for some cases. This suggests that we want it in 4.0. It is also an output change for other cases, like going from "8M" to "8 MiB". We probably can't tell for sure whether some tools expect the spelling "8M" (even if this is supposed to be the human interface and tools should be using JSON) or feed the change back to qemu-img or qemu-io (which accept "8M", but not "8 MiB" as sizes in most places). This suggests that we shouldn't make this change as late as -rc2. So what is the conclusion? Kevin
