Alternative download paths for RSB?

[Chris Johns]

On 22/12/21 1:45 am, Joel Sherrill wrote:
> On Tue, Dec 21, 2021 at 12:33 AM Sebastian Huber
> <sebastian.huber at embedded-brains.de> wrote:
>>
>> On 20/12/2021 22:52, Chris Johns wrote:
>>> On 21/12/21 2:33 am, Joel Sherrill wrote:
>>>> On Mon, Dec 20, 2021 at 8:54 AM Joel Sherrill <joel at rtems.org> wrote:
>>>>>
>>>>> On Mon, Dec 20, 2021 at 6:59 AM Sebastian Huber
>>>>> <sebastian.huber at embedded-brains.de> wrote:
>>>>>>
>>>>>> Hello,
>>>>>>
>>>>>> it seems the GNU web server is down:
>>>>>
>>>>> Yep. Still down.
>>>>>
>>>>> i reached out to the sourceware admins to see if they knew who to
>>>>> contact. I couldn't find a gnu.org address.
>>>>
>>>> Social media post from FSF forwarded by one of the sourceware admins
>>>> about a general outage at the FSF:
>>>> https://hostux.social/@fsfstatus/107479476196064925
>>>>
>>>> Looks to be up now. RSB handling mirrors would be good. Although
>>>> it would seem to be complex to ensure that all entries and mirrors are
>>>> correct.
>>>>
>>>>>
>>>>>>
>>>>>> download: https://ftp.gnu.org/gnu/mpfr/mpfr-3.1.4.tar.bz2 ->
>>>>>> sources/mpfr-3.1.4.tar.bz2
>>>>>> download: https://ftp.gnu.org/gnu/mpfr/mpfr-3.1.4.tar.bz2: error:
>>>>>> <urlopen error [Errno 101] Network is unreachable>
>>>>>> error: downloading https://ftp.gnu.org/gnu/mpfr/mpfr-3.1.4.tar.bz2: all
>>>>>> paths have failed, giving up
>>>>>> Build FAILED
>>>>>> error: downloading https://ftp.gnu.org/gnu/mpfr/mpfr-3.1.4.tar.bz2: all
>>>>>> paths have failed, giving up
>>>>>> Build Set: Time 0:04:23.318509
>>>>>> Build FAILED
>>>>>>
>>>>>> The error message suggests that it is possible to add alternative
>>>>>> download sources. How can I do this? This didn't work:
>>>>>>
>>>>>> diff --git a/source-builder/config/gcc-9.cfg
>>>>>> b/source-builder/config/gcc-9.cfg
>>>>>> index f8cf69b..1d7c716 100644
>>>>>> --- a/source-builder/config/gcc-9.cfg
>>>>>> +++ b/source-builder/config/gcc-9.cfg
>>>>>> @@ -21,4 +21,8 @@
>>>>>>    %source set gmp https://ftp.gnu.org/gnu/gmp/gmp-%{gmp_version}.tar.bz2
>>>>>>    %source set mpfr https://ftp.gnu.org/gnu/mpfr/mpfr-%{mpfr_version}.tar.bz2
>>>>>>
>>>>>> +%source set mpc
>>>>>> https://ftp.mpi-inf.mpg.de/pub/mirrors/gnu/mirror/gcc.gnu.org/pub/gcc/infrastructure/mpc-%{mpc_version}.tar.gz
>>>>>> +%source set gmp
>>>>>> https://ftp.mpi-inf.mpg.de/pub/mirrors/gnu/mirror/gcc.gnu.org/pub/gcc/infrastructure/gmp-%{gmp_version}.tar.bz2
>>>>>> +%source set mpfr
>>>>>> https://ftp.mpi-inf.mpg.de/pub/mirrors/gnu/mirror/gcc.gnu.org/pub/gcc/infrastructure/mpfr-%{mpfr_version}.tar.bz2
>>>>>> +
>>>>>>    %include %{_configdir}/gcc-common-1.cfg
>>>>>
>>>>> This would be Chris' department. Sorry
>>>
>>> I would update the `%source` keyword to accept:
>>>
>>> %source mirror mpc 1 https://ftp.mpi-inf.mpg.de/../mpc-%{mpc_version}.tar.gz
>>>
>>> where `1` is an integer priority and 1 is highest. The code is here:
>>>
>>> https://git.rtems.org/rtems-source-builder/tree/source-builder/sb/sources.py#n55
>>>
>>> There is no such thing as a list in the macros so I suspect you would need to
>>> dereference the source where the results of `_make_key` would reference another
>>> series of macro keys that are the mirrors.
>>>
>>> Do we want to start maintaining lists of mirrors in RSB configurations? I am
>>> undecided. As developers we do not notice the online status due to the RSB
>>> cache. For sites and users would deployment be a better long term solution?
>>
>> Thanks for the hints. I added a ticket for this:
>>
>> https://devel.rtems.org/ticket/4573
> 
> How can we ensure that all mirrors listed work?

They would need to be tested and that means more code in the RSB to do it.

Chris