From: Brandon Casey <casey@nrlssc.navy.mil>
Date: Wed, 16 Jan 2008 20:46:23 GMT
Subject: Re: [PATCH 2/2] close_lock_file(): new function in the lockfile API
Message-ID: <Pine.LNX.4.64.0801161443340.31161@torch.nrlssc.navy.mil>
In-Reply-To: <7vodblo6c9.fsf@gitster.siamese.dyndns.org>

On Wed, 16 Jan 2008, Junio C Hamano wrote:

> Linus Torvalds <torvalds@linux-foundation.org> writes:
>
>> On Wed, 16 Jan 2008, Junio C Hamano wrote:
>>> +
>>> +void close_lock_file(struct lock_file *lk)
>>> +{
>>> +	close(lk->fd);
>>> +	lk->fd = -1;
>>> +}
>>
>> Since one of the main purposes of closing would be the error testing of
>> writes that haven't made it out yet on filesystems like NFS that do
>> open-close cache serialization, I'd suggest doing this as
>>
>> 	int close_lock_file(struct lock_file *lk)
>> 	{
>> 		int fd = lk->fd;
>> 		lk->df = -1;
>> 		return close(fd);
>> 	}
>>
>> to give the return code.
>
> Yup!  You are as always right.

My patch does this, though I understand it may take some time to review.

I left the lk->fd unmodified when close() failed in case the caller
would like to include it in an error message.

-brandon