Jac Gl wrote:
> Hi,
> I am dissapointed as well. I found something like this:
> This new build system was created in general to get reliable 'stable' daily builds of tvheadend for ubuntu for all people. Having said that, the dockerhub build farm has only intel
Sure that's wrong. I should correct that to say 'most people', or 'a stable public builds' instead of 'all people'. Because not all hardware can be supported. Only the most common ones.
Jac Gl wrote:
> machines. Therefore it can only produce x86 .deb files. And the build scripts are only making 64-bit of type x86_64. As I don't really expect that there are enough people still requiring i386 for 32-bit to be worth making anymore."
> Hmmm... isn't it good idea to utilize old PC as TV server, without display/keyboard/mouse? I have the one with Pentium4, 3 DVB-T cards and DVB-S2 card, and to be honest there is no need to have better cpu. Tvheadend takes maybe 10-20% of proc time.
> Regards.
Yes you are correct in that philosophy, however except that in environmental terms specifically the idle power consumption of the Pentium4 CPU isn't a terribly good match if it's idle TDP is so high (including ATX psu and other attached components), when based on such old process technology.
My main point about cutting the line at i386 is because before the era of Core2Duo architecture, most CPUs such as P4 were not anywhere near as power effecicient. And hence it makes a good place to say: well how many years of support (from 2015 onward) to you expect? Can you put a number on that in years,months ?
Saen Acro has a ULV BGA part which is very low and reasonable TDP - so in his specific case it's a pretty good match for use as Tvheadend. Which is a good example of where an i386 build is justified. Yet expect the total number of active systems purchased with Core2Duo part during those era dwarfs such (most likely ULV laptop).
So that's all good justification to hear. Yet the practical side is different. The system just cannot build every possible architecture because there are not unlimited resources.
These builds are done on public machines. Therefore i386 builds are not made to avoid jeopardizing other more important builds going on at the same time when the system becomes overloaded.
e.g. 3 users << 100 users
The numbers just don't add up.
* We need to put priority ahead for ARM builds - because that is more users.
* We need to put priority ahead for RPM builds - same reason (*x64 only RPM)
This is a conservative approach. Whereby we act in a responsible fashion and cannot afford to build everything, else risk failed builds.
For example, my inbox this morning:
notifications@docker.com
Hi Dreamcat4,
There seems to have been an issue with your Automated Build "dreamcat4/dpi" (VCS repository: dreamcat4/dpi) during the build step.
You can find more information on
https://registry.hub.docker.com/u/dreamcat4/dpi/build_id/61282/code/bw38stba8sny4k5bjx7erhu/
... that's just the current situation with only 64-bits build. Master didn't get updated last night due to over-use of builds farm.
We continue to monitor this. It might be in future that:
* Docker Builds farm gets upgraded, there becomes more resources.
* Docker Builds farm gets more popular, there becomes less resources.
We cannot predict the future situation with Docker Hub... it is something entirely out of our control.
What about other servers?
Launchpad.net
* Native git hosting will be upgraded in 'unknown months time'. The people working on it have not set any schedule whatsoever.
* Until then no tvheadend PPA can be build on launchpad.net due to bzr launchpad bug regarding git submodules (bzr git import fails).
* I have not got any time allocated to make a brand new launchpad auto-builds. It might have to be somebody else to do it.
I hope that clarifies the situation for you guys in terms of why i386 builds are not being made. There was no official policy decision. Given unlimited resources we would have tried to make as many other (lower demanded) build as possible.
So it is a very good thing that Saen Acro is making here community i386 builds and sharing them publically. I support that effort, (just like Nihil who make rpi ARM build himself currently). It is a community effort.
