qemu-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Qemu-devel] [PATCH 02/21] qcow2: Add refcount_width to format-speci


From: Eric Blake
Subject: Re: [Qemu-devel] [PATCH 02/21] qcow2: Add refcount_width to format-specific info
Date: Tue, 11 Nov 2014 08:49:23 -0700
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:31.0) Gecko/20100101 Thunderbird/31.2.0

On 11/11/2014 01:11 AM, Max Reitz wrote:
>>> +            .compat             = g_strdup("0.10"),
>>> +            .refcount_width     = s->refcount_bits,
>> Hmm - is it really worth displaying a constant?  Since the 0.10 format
>> cannot change the width from 16, I'm not sure if it adds anything to the
>> output to display it.  After all, there's other things we omit for the
>> old format when they cannot be altered (such as the state of a lazy
>> flag).  On the other hand, if it makes your changes to later iotests
>> easier for tests that operate on both image formats, I'm not opposed
>> to it.
> 
> Yes, I thought about not displaying it. But whereas "corrupt" or "lazy
> refcounts" simply do not make sense with compat=0.10 images (it's simply
> impossible), the refcount width does make sense. It's always 16 bits
> (I'm noticing myself how I keep swapping between "bit" and "bits", but I
> just can't help it) but I personally find it interesting enough to
> display. I'd be fine with dropping it from compat=0.10, though.
> 
> But in retrospect, I'd rather make the other two flags always visible
> than now drop this entry. However, not displaying a bool if it's always
> false makes more sense to me than not displaying an integer because it's
> always constant.
> 

>> If you can make a strong argument for always outputting the constant
>> width of 16 for 0.10 formats, then I can live with it, so:
> 
> You decide whether it's strong enough. :-)
> 
> My main argument is "If a bool is not displayed one can assume it to be
> false; if an integer is not displayed which naturally cannot be 0, I
> will have no idea what it would be, even if it's constant for that image
> version".

Sounds fairly convincing :)  Add a paragraph like that to the commit
message, and I'm sold!

> 
>> Reviewed-by: Eric Blake <address@hidden>

So looks like you get to keep this.

-- 
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org

Attachment: signature.asc
Description: OpenPGP digital signature


reply via email to

[Prev in Thread] Current Thread [Next in Thread]