To: Martin Jansa <martin.jansa@gmail.com>
References: <56E174E6.1050305@mlbassoc.com> <20160310135128.GH2573@jama>
From: Gary Thomas <gary@mlbassoc.com>
Message-ID: <56E17BD7.1070503@mlbassoc.com>
Date: Thu, 10 Mar 2016 14:51:19 +0100
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:38.0) Gecko/20100101
	Thunderbird/38.5.1
MIME-Version: 1.0
In-Reply-To: <20160310135128.GH2573@jama>
Cc: yocto@yoctoproject.org
Subject: Re: Dualing downloads
Precedence: list
Content-Type: text/plain; charset=windows-1252; format=flowed
Content-Transfer-Encoding: 7bit

On 2016-03-10 14:51, Martin Jansa wrote:
> On Thu, Mar 10, 2016 at 02:21:42PM +0100, Gary Thomas wrote:
>> If I have a package X that depends on X-native and they build
>> from the same sources, doesn't the fetch step duplicate effort
>> if the sources need to be downloaded from scratch?  I just saw
>> this [appear to] happen and it's not clear if the sources were
>> downloaded once or twice.
>
> The first one should create the lock file and second fetch task should
> just reuse the already downloaded source.
>
> Otherwise sharing whole downloads directory between multiple builders
> wouldn't work as well.
>

That's great, thanks

>> I this is the case, perhaps if X depends on X-native then X.fetch
>> should depend on X-native.fetch so then download happens only once?
>> Or is that too hard to detect?
>>
>> For many packages (and fast network links) this probably isn't an
>> issue, but I could imagine some very large packages simply wasting
>> network bandwidth.
>>
>> Note: I'm happy if I'm wrong about this, it just jumped out at me.

-- 
------------------------------------------------------------
Gary Thomas                 |  Consulting for the
MLB Associates              |    Embedded world
------------------------------------------------------------