[pacman-dev] [PATCH] libalpm: download rates becoming negative

[Morgan Adamiec]

On 19/04/2021 17:08, Anatol Pomozov wrote:
> Hi
> 
> On Mon, Apr 19, 2021, 02:46 Morgan Adamiec <morganamilo at archlinux.org>
> wrote:
> 
>>
>>
>> On 19/04/2021 13:39, Andrew Gregory wrote:
>>> On 04/19/21 at 01:32pm, Morgan Adamiec wrote:
>>>>
>>>>> Because the file may not be the same after we fallback to a different
>>>>> server.
>>>>>
>>>>
>>>> That shouldn't be the case though. Do we actually support that use case?
>>>
>>> Yes.  Databases.
>>>
>>>>>> In fact I don't see the point of this variable at all. Why would we
>> ever
>>>>>> not want to unlink on fail?
>>>>>
>>>>> So that it can be resumed later.
>>>>>
>>>>
>>>> I guess, though the above point applies. it may be a different mirror
>>>> and therefore the file may not be the same. So I feel if you can resume
>>>> between pacman calls you should be able to resume between servers.
>>
>> Ah yeah I guess databases makes sense. But why not just have
>> allow_resume as false for those downloads and allow package downloads to
>> continue?
>>
> 
> If it cannot update the packages database the there is not much sense to
> start downloading packages.
> 
>>


This is in curl_retry_next_server. It's about not clearing the progress
when switching from one server to the next. A mirror may fail to
download the database but overall it can still succeed and continue to
package download.