Re: Removing armel from the list of arches we build for
2013/8/21 Joey Hess <email@example.com>:
> Gustavo Noronha Silva wrote:
>> armel has been one of the sore points of maintaining WebKit in Debian.
>> It often fails to build the package because of lack of memory or disk
>> space. I'm told by my colleague Hector Oron that WebKit's utility for
>> armel is quite limited, given it's often used headless rather than
>> providing a graphical user interface of some kind.
> Sounds like the same kind of "surely noone uses this" rationalization we
> always see when going back to see why some package we use got removed
> from Debian.
The fundamental cause for the build failure seems to be memory
limitation, while webkit requires all the available RAM in buildd
(1.5GB), is uses all the SWAP (3GB) and final linking stage explodes
only after 3d 14h 13m or so ... we could still try to make swap
bigger and re-try the build, but certainly we are reaching buildd
hardware limit here. This is not the lame excuse of "surely noone uses
this" rationalization we always see because someone does not feel like
wasting time on fixing a bug on non his preferred architecture.
However, more powerful hardware is now commercially available, Debian
recently got donated 3 nodes on Calxeda servers with Quad core
Cortex-A9 with 4GB of RAM, however those were planned to be used for
armhf port (maybe as two buildds and one porterbox), and current
infrastructure does not easily allow to special case package builds
Having said the above, we are still working on trying to fix the build
of webkitgtk/armel, but no promises here.
2013/8/21 peter green <firstname.lastname@example.org>:
> The bigger question IMO though is how many reverse dependencies does the
> variant of webkit built by the "webkit" source package (which is only one of
> several webkit variants in debian) have and how strong are those reverse
Héctor Orón -.. . -... .. .- -. -.. . ...- . .-.. --- .--. . .-.