Re: [Qemu-devel] [PATCH v14 13/15] qapi: Enforce (or whitelist) case conventions on qapi members

[Markus Armbruster]

Eric Blake <address@hidden> writes:

> On 12/02/2015 10:19 AM, Markus Armbruster wrote:
>
>>> With your patch, the positive tests no longer work in isolation.  You
>>> were getting lucky that things sorted such that 'Foo' was checked for
>>> correctness prior to 'UuidInfo'; but if you comment out the 'Foo'
>>> declaration, or rename from 'Foo' to something else that hashes after
>>> 'UuidInfo', then args-member-case and union-branch-case start reporting
>>> failures about UuidInfo (and only enum-member-case honors the
>>> whitelist).
>> 
>> You're right.
>> 
>
>>> +++ w/tests/qapi-schema/args-member-case.json
>>> @@ -1,3 +1,3 @@
>>>  # Member names should be 'lower-case' unless the struct/command is
>>> whitelisted
>>> -{ 'command': 'UuidInfo', 'data': { 'Arg': 'int' } }
>>> +{ 'command': 'x-UuidInfo', 'data': { 'Arg': 'int' } }
>> 
>> Will fail as soon as we enforce the command naming convention.  To avoid
>> that, we need to add a suitable name to the whitelist.
>> 
>> However, we don't actually have any command parameters to whitelist.
>> Why bother testing it then?
>
> Good point.
>
>
>> 
>>>From 44f07a40c8b9b5d1f24833028b5dacde1fd50c80 Mon Sep 17 00:00:00 2001
>> From: Markus Armbruster <address@hidden>
>> Date: Wed, 2 Dec 2015 17:41:55 +0100
>> Subject: [PATCH] fixup! qapi: Enforce (or whitelist) case conventions on qapi
>>  members
>> 
>> Signed-off-by: Markus Armbruster <address@hidden>
>> ---
>>  scripts/qapi.py                          | 6 ++----
>>  tests/qapi-schema/args-member-case.err   | 2 +-
>>  tests/qapi-schema/args-member-case.json  | 3 +--
>>  tests/qapi-schema/enum-member-case.err   | 2 +-
>>  tests/qapi-schema/enum-member-case.json  | 4 ++--
>>  tests/qapi-schema/union-branch-case.err  | 2 +-
>>  tests/qapi-schema/union-branch-case.json | 3 +--
>>  7 files changed, 9 insertions(+), 13 deletions(-)
>
> Looks good.  Commit 6c1eb345 on your qapi-not-next branch has this
> already squashed in, but lacking your S-o-b.  But that only affects the
> commit message, so I'll go ahead and base my future patch submissions on
> top of the current contents of qapi-not-next (it will be interesting to
> see how many patches land from various trees right after 2.6 opens up...)

I intended to push with a separate fixup commit; looks like a last
minute rebase to tweak commit messages squashed it in.  Anyway, I fixed
up the commit message and pushed again.  Thanks!