Re: [9front] Mixfs: potential bug?

[Trevor Higgins <plan9fullfrontal@qs.co.nz>]

Ok, thanks. Sorry for being too lazy to just compile some code and check 
it out my self. Which I have done now.

I got the first part of anon structures long ago but missed the second 
point. about compiler automunging the references.

I had a crack at porting hjfs to front9port, by naming the anon 
structures and converting the refs to &struct->qlock et al

but it was not 100% successful.

The program runs without seg faulting but does not create 
'namespace'/hjfs.cmd and hangs somewhere without serving anthing on the 
mount point.

I will now have to check that I put the correct structure reference into 
the edits.


Thanks again.


On 02/13/2020 01:21 PM, Silas McCroskey wrote:
> http://doc.cat-v.org/plan_9/4th_edition/papers/comp
>
> The "Extensions" section goes over this compiler extension:
>
> structs-within-structs declared without a name allow their members to
> be accessed directly, so
> qlock(s)
> does the same thing as
> qlock(s->Lock)
> since the "locked" member used by qlock is accessible to both.
>
> This is a compiler extension rather than pointer tricks, so it does
> not depend on ordering within the struct (as displayed in the
> Extensions section itself, where Lock is the last member of the outer
> struct).
>
> - sam-d
>
> On Wed, Feb 12, 2020 at 4:07 PM Trevor Higgins
> <plan9fullfrontal@qs.co.nz> wrote:
>> Looking at the code for audio/mixfs (mixfs.c)
>>
>> The code uses qlock(s) where s is: struct Stream*.
>>
>> In the structure ,shouldn't the QLock member of the structure be the
>> first member in the structure?
>> If not so, then why does it not matter?
>>
>> Thanks.
>>
>> --
>>
-- 
We need another plan