On Wed, 4 Jan 2012 17:38:20 +0000 Stefan Hajnoczi <stefa...@linux.vnet.ibm.com> wrote:
> The block resize command returns undefined errors if things go wrong. This is > bad since users will have no chance to understand what failed. Oops, this introduces a segfault try "block_resize ide1-cd0 100" in the monitor and you'll see it. > > Patch 3 makes qmp_block_resize() use meaningful errors. We introduce new > qerrors for ENOMEDIUM and EACCES since nothing exists yet. > > When doing this I noticed that qerror definitions and table entries are not in > alphabetical order as required by the comments in qerror.h and qerror.c. > That's not a surprise since there is no automated way to enforce this. I'm > not > sure whether this requirement is useful or not in the first place, but felt > guilty creating more mess. So I've restored alphabetical order and added a > script to verify that the requirement is met in Patches 1 and 2. > > Stefan Hajnoczi (3): > qerror: add check-qerror.sh to verify alphabetical order > qerror: restore alphabetical order over qerrors > block: use proper qerrors in qmp_block_resize > > blockdev.c | 26 +++++++++---- > qerror.c | 91 +++++++++++++++++++++++++--------------------- > qerror.h | 78 +++++++++++++++++++++------------------ > scripts/check-qerror.sh | 22 +++++++++++ > 4 files changed, 131 insertions(+), 86 deletions(-) > create mode 100755 scripts/check-qerror.sh >
