The following changes since commit 19a720b74fde7e859d19f12c66a72e545947a657:

  Merge tag 'tracing-pull-request' of https://gitlab.com/stefanha/qemu into staging (2023-06-01 08:30:29 -0700)

are available in the Git repository at:

  https://repo.or.cz/qemu/ericb.git tags/pull-nbd-2023-06-01

for you to fetch changes up to 58516caac47c4bc7ed3ad6a8e2f565404a563dd3:

  cutils: Improve qemu_strtosz handling of fractions (2023-06-01 16:55:25 -0500)

----------------------------------------------------------------
nbd and misc patches for 2023-06-01

- Eric Blake: Fix iotest 104 for NBD
- Eric Blake: Improve qcow2 spec on padding bytes
- Eric Blake: Fix read-beyond-bounds bug in qemu_strtosz

----------------------------------------------------------------
Eric Blake (21):
      iotests: Fix test 104 under NBD
      qcow2: Explicit mention of padding bytes
      test-cutils: Avoid g_assert in unit tests
      test-cutils: Use g_assert_cmpuint where appropriate
      test-cutils: Test integral qemu_strto* value on failures
      test-cutils: Test more integer corner cases
      cutils: Fix wraparound parsing in qemu_strtoui
      cutils: Document differences between parse_uint and qemu_strtou64
      cutils: Adjust signature of parse_uint[_full]
      cutils: Allow NULL endptr in parse_uint()
      test-cutils: Add coverage of qemu_strtod
      test-cutils: Prepare for upcoming semantic change in qemu_strtosz
      test-cutils: Refactor qemu_strtosz tests for less boilerplate
      cutils: Allow NULL str in qemu_strtosz
      numa: Check for qemu_strtosz_MiB error
      test-cutils: Add more coverage to qemu_strtosz
      cutils: Set value in all qemu_strtosz* error paths
      cutils: Set value in all integral qemu_strto* error paths
      cutils: Use parse_uint in qemu_strtosz for negative rejection
      cutils: Improve qemu_strtod* error paths
      cutils: Improve qemu_strtosz handling of fractions

 docs/interop/qcow2.txt           |    1 +
 include/qemu/cutils.h            |    5 +-
 audio/audio_legacy.c             |    4 +-
 block/gluster.c                  |    4 +-
 block/nfs.c                      |    4 +-
 blockdev.c                       |    4 +-
 contrib/ivshmem-server/main.c    |    4 +-
 hw/core/numa.c                   |   11 +-
 qapi/opts-visitor.c              |   10 +-
 tests/unit/test-cutils.c         | 2469 ++++++++++++++++++++++++++++----------
 ui/vnc.c                         |    4 +-
 util/cutils.c                    |  263 ++--
 util/guest-random.c              |    4 +-
 util/qemu-sockets.c              |   10 +-
 tests/qemu-iotests/common.filter |    4 +-
 tests/qemu-iotests/common.rc     |    3 +-
 tests/qemu-iotests/049.out       |    7 +-
 tests/qemu-iotests/178.out.qcow2 |    3 +-
 tests/qemu-iotests/178.out.raw   |    3 +-
 19 files changed, 2035 insertions(+), 782 deletions(-)

-- 
2.40.1

On 6/1/23 15:02, Eric Blake wrote:
> Eric Blake (21):
>        iotests: Fix test 104 under NBD
>        qcow2: Explicit mention of padding bytes
>        test-cutils: Avoid g_assert in unit tests
>        test-cutils: Use g_assert_cmpuint where appropriate
>        test-cutils: Test integral qemu_strto* value on failures
>        test-cutils: Test more integer corner cases
>        cutils: Fix wraparound parsing in qemu_strtoui
>        cutils: Document differences between parse_uint and qemu_strtou64
>        cutils: Adjust signature of parse_uint[_full]
>        cutils: Allow NULL endptr in parse_uint()
>        test-cutils: Add coverage of qemu_strtod
>        test-cutils: Prepare for upcoming semantic change in qemu_strtosz
>        test-cutils: Refactor qemu_strtosz tests for less boilerplate
>        cutils: Allow NULL str in qemu_strtosz
>        numa: Check for qemu_strtosz_MiB error
>        test-cutils: Add more coverage to qemu_strtosz
>        cutils: Set value in all qemu_strtosz* error paths
>        cutils: Set value in all integral qemu_strto* error paths
>        cutils: Use parse_uint in qemu_strtosz for negative rejection
>        cutils: Improve qemu_strtod* error paths
>        cutils: Improve qemu_strtosz handling of fractions

This is failing on Windows (32 and 64-bit):

https://gitlab.com/qemu-project/qemu/-/jobs/4399466166#L3524
https://gitlab.com/qemu-project/qemu/-/jobs/4399466165#L3332

|  21/135 /cutils/qemu_strtol/overflow - 
ERROR:../tests/unit/test-cutils.c:1387:test_qemu_strtol_overflow: assertion failed (res == 
LONG_MIN): (2147483647 == -2147483648) FAIL

It seems to have returned LONG_MAX instead of LONG_MIN.


r~

On Thu, Jun 01, 2023 at 08:58:46PM -0700, Richard Henderson wrote:
> On 6/1/23 15:02, Eric Blake wrote:
> > Eric Blake (21):
> >        iotests: Fix test 104 under NBD
> >        qcow2: Explicit mention of padding bytes
> >        test-cutils: Avoid g_assert in unit tests
> >        test-cutils: Use g_assert_cmpuint where appropriate
> >        test-cutils: Test integral qemu_strto* value on failures
> >        test-cutils: Test more integer corner cases
> >        cutils: Fix wraparound parsing in qemu_strtoui
> >        cutils: Document differences between parse_uint and qemu_strtou64
> >        cutils: Adjust signature of parse_uint[_full]
> >        cutils: Allow NULL endptr in parse_uint()
> >        test-cutils: Add coverage of qemu_strtod
> >        test-cutils: Prepare for upcoming semantic change in qemu_strtosz
> >        test-cutils: Refactor qemu_strtosz tests for less boilerplate
> >        cutils: Allow NULL str in qemu_strtosz
> >        numa: Check for qemu_strtosz_MiB error
> >        test-cutils: Add more coverage to qemu_strtosz
> >        cutils: Set value in all qemu_strtosz* error paths
> >        cutils: Set value in all integral qemu_strto* error paths
> >        cutils: Use parse_uint in qemu_strtosz for negative rejection
> >        cutils: Improve qemu_strtod* error paths
> >        cutils: Improve qemu_strtosz handling of fractions
> 
> This is failing on Windows (32 and 64-bit):
> 
> https://gitlab.com/qemu-project/qemu/-/jobs/4399466166#L3524
> https://gitlab.com/qemu-project/qemu/-/jobs/4399466165#L3332
> 
> |  21/135 /cutils/qemu_strtol/overflow -
> ERROR:../tests/unit/test-cutils.c:1387:test_qemu_strtol_overflow: assertion
> failed (res == LONG_MIN): (2147483647 == -2147483648) FAIL
> 
> It seems to have returned LONG_MAX instead of LONG_MIN.

Gah.  I see the problem; it is a copy/paste typo in part of
test-cutils.c guarded by 'if (LONG_MAX == INT_MAX)' and therefore
doesn't fire on platforms with 64-bit long.  Will respin the pull
request to fix it.

-- 
Eric Blake, Principal Software Engineer
Red Hat, Inc.           +1-919-301-3266
Virtualization:  qemu.org | libvirt.org

Pattern:

First, one of us gets a bright idea on user-friendly interface (here:
fractional sizes like 1.5M).  Objections, if any, get brushed aside.

Then the thing sprouts warts, tentacles, sores, and starts to give off
that sickly-sweet smell of bugs feasting on misguided ideas.

Until one of us spends a lot more time on containing and reducing the
damage than the thing could ever be worth.

Cobbler, stick to your last.

Thanks, Eric!

On Fri, Jun 02, 2023 at 08:32:34AM +0200, Markus Armbruster wrote:
> Pattern:
> 
> First, one of us gets a bright idea on user-friendly interface (here:
> fractional sizes like 1.5M).  Objections, if any, get brushed aside.
> 
> Then the thing sprouts warts, tentacles, sores, and starts to give off
> that sickly-sweet smell of bugs feasting on misguided ideas.
> 
> Until one of us spends a lot more time on containing and reducing the
> damage than the thing could ever be worth.
> 
> Cobbler, stick to your last.
> 
> Thanks, Eric!

The scary part was that I found several other bugs completely
unrelated to the read-out-of-bounds, merely by increasing unit test
coverage.  When Hanna first pointed me to the problem asking if a
5-line patch would work, I was not expecting it to blow up into a
series touching more than 1000 lines (true, most of those lines were
in the unit tests).

-- 
Eric Blake, Principal Software Engineer
Red Hat, Inc.           +1-919-301-3266
Virtualization:  qemu.org | libvirt.org

Eric Blake <eblake@redhat.com> writes:

> The scary part was that I found several other bugs completely
> unrelated to the read-out-of-bounds, merely by increasing unit test
> coverage.  When Hanna first pointed me to the problem asking if a
> 5-line patch would work, I was not expecting it to blow up into a
> series touching more than 1000 lines (true, most of those lines were
> in the unit tests).

You spot vermin hiding under a rock.  What will you likely see when you
lift the rock?

Thanks again for lifting the rock :)