On Thu, 3 Nov 2016 12:19:56 -0700 Khem Raj <[email protected]> wrote:
> > > On Nov 3, 2016, at 6:06 AM, Gary Thomas <[email protected]> wrote: > > > > I recall seeing some discussion in the past about using shallow > > GIT clones when importing repositories? Is this ever going to > > happen? > > > > The reason I ask is that I routinely save the GIT tarballs and > > some of them are obscenely obese :-( The worst of the bunch > > is the device firmware for the RaspberryPi (as of today): > > -rw-rw-r-- 1 gthomas gthomas 6321877646 Nov 3 12:13 > > git2_github.com.raspberrypi.firmware.git.tar.gz > > > > This particular tar file increased by more than 300MB since > > the last time I downloaded it (only 2016-09-13!) I routinely > > slosh these files across the oceans (sometimes using tin cans > > and strings it seems) and this can be very tedious. Is there > > anything that can be done to make these files a bit more manageable? > > while shallow clones is a comprehensive solution and we should probably slot > it for 2.3 release > the above recipe should stop using git fetcher and convert to using tarballs > since this git repo hosts binaries, it will bloat with every time they push > stuff into it. Someone should teach these RPi folks to not abuse github. > I raised this years ago: https://github.com/raspberrypi/firmware/issues/175 I also did a bit of investigation and found that the files we actually use from the firmware repo total under 3MB as a tar.xz file. I have scripts which can create an appropriate tar.xz file from the git repo and so I could start publishing these archives again if there's any interest in using them in the meta-raspberrypi layer. At 3MB per archive, even if we update the firmware once per week this would never get as bad as the current firmware repository. Thanks, Paul Barker -- _______________________________________________ yocto mailing list [email protected] https://lists.yoctoproject.org/listinfo/yocto
