On Thu, 2008-08-14 at 14:27 -0400, Matthew Miller wrote: > So, I'm updating BU Linux 5 (CentOS5 based) to use yum 3.2.18, and > discovered a perplexing problem: On a newly-installed system with no rpm gpg > keys imported, rather than offering to import the gpgkey defined in the repo > file, it tells me that I ought to define one there (or else run rpm --import > manually on a filename which does not exist -- a confusing error message if > I didn't know better). > > In playing around, I discovered that if I import a gpg key, run yum, and > then remove the key with rpm -e, yum works just fine -- it asks to import > the key as it ought. > > So, clearly, yum's saving some sort of state (in an unhelpful way). A little > investigation turns up /var/cache/yum/.gpgkeyschecked.yum -- touching that > file removes the problem. > > So, I went to yum/__init__.py and patched the gpgKeyCheck function to always > return 1 (as if the .gpgkeyschecked.yum file is always there) and everything > seems great. > > So, uh, first of all: Buh? What? There seem to be two code paths for > checking if there are gpg keys, the working one which imports the keys, and > this function which is all, like, broke and stuff. > > Second, assuming that gets fixed, the message telling people to run "rpm > --import public.gpg.key" needs to be changed, because running that command > will cause people to report this error: > > error: public.gpg.key: import read failed(-1). > > which helps no one.
We're importing the keys to another location now, in addition to the rpm database. I must have messed up something on the import check. Can you recreate this consistently? -sv _______________________________________________ Yum-devel mailing list [email protected] https://lists.dulug.duke.edu/mailman/listinfo/yum-devel
