On Mon, Jun 20, 2016 at 10:01 AM, Anton Korobeynikov via llvm-dev < llvm-dev at lists.llvm.org> wrote:> > P.S.: On a similar note, are there any news regarding llvm.org/apt? > We are working on it. Note, however, that it seems that the majority > of bogus load seemed to come from CI systems, which pulled apt repo > for every and each downstream commit without any caching / checking > whether the mainline changed. We would certainly try to limit such > behavior if / when new APT mirror will be established. >I believe many projects have now switched to uncached downloads from llvm.org/releases (Boost.Hana also does this uncached atm.). Caching on e.g. Travis and CircleCI is not easily available for APT repositories; I think that's why noone bothered. For simple file downloads it is much easier, so perhaps a little explaining text can be added to the message on llvm.org/apt . Example Travis script with caching: https://github.com/ldc-developers/ldc/blob/master/.travis.yml -Johan -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20160620/2234db49/attachment.html>
On 20 June 2016 at 10:17, Johan Engelen via llvm-dev <llvm-dev at lists.llvm.org> wrote:>> > P.S.: On a similar note, are there any news regarding llvm.org/apt? >> We are working on it. Note, however, that it seems that the majority >> of bogus load seemed to come from CI systems, which pulled apt repo >> for every and each downstream commit without any caching / checking >> whether the mainline changed. We would certainly try to limit such >> behavior if / when new APT mirror will be established. > > I believe many projects have now switched to uncached downloads from > llvm.org/releases (Boost.Hana also does this uncached atm.). Caching on e.g. > Travis and CircleCI is not easily available for APT repositories; I think > that's why noone bothered.That's a poor excuse. :)> For simple file downloads it is much easier, so > perhaps a little explaining text can be added to the message on llvm.org/aptWell, adding a README doesn't mean people will read or worse, abide. We need to be more pro-active on defending our infrastructure, so I suggest using Apache's mod_evasive. cheers, --renato
On Mon, Jun 20, 2016 at 11:40 AM, Renato Golin <renato.golin at linaro.org> wrote:> On 20 June 2016 at 10:17, Johan Engelen via llvm-dev > <llvm-dev at lists.llvm.org> wrote: > >> > P.S.: On a similar note, are there any news regarding llvm.org/apt? > >> We are working on it. Note, however, that it seems that the majority > >> of bogus load seemed to come from CI systems, which pulled apt repo > >> for every and each downstream commit without any caching / checking > >> whether the mainline changed. We would certainly try to limit such > >> behavior if / when new APT mirror will be established. > > > > I believe many projects have now switched to uncached downloads from > > llvm.org/releases (Boost.Hana also does this uncached atm.). Caching on > e.g. > > Travis and CircleCI is not easily available for APT repositories; I think > > that's why noone bothered. > > That's a poor excuse. :) > >Please show me the travis.yml or circle.yml file that does it ;-) - Johan -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20160620/e7329a31/attachment.html>
Johan Engelen via llvm-dev <llvm-dev <at> lists.llvm.org> writes:> [..] > > I believe many projects have now switched to uncached downloads from > llvm.org/releases (Boost.Hana also does this uncached atm.).I've just implemented caching for LLVM downloads in Hana's CI. I'm testing it and I'll merge it as soon as it's fully working.