[Rpm-maint] Is there a proper way how to get byte range of header?

Ales Kozumplik akozumpl at redhat.com
Thu Jun 5 08:33:49 UTC 2014


On 06/04/2014 02:08 PM, Panu Matilainen wrote:
> Does something still use that data? AFAIK the header byte range in
> repodata is nothing but a hack which yum needed at dawn of times, but
> long since unneeded.

I think Yum in Anaconda still needs it: to save disk space it does not 
download all the RPMs but starts transaction without them and then 
downloads them one by one as RPM requests them through callback. It 
still needs to feed RPM something at the start though and so it uses the 
headers. Is there a better way?

The only reason DNF does not does the same yet is that we use parallel 
downloads and so herd all the RPMs before transation starts. But it can 
happen someone will run into an installation where all the RPMs won't 
fit on the disk and we'll have to implement what Yum does.

Ales


More information about the Rpm-maint mailing list