> > Judging from the ipkg-push file, it's made to support not just opkg, but > > also ipkg, which I don't think supports the SHA256 checksum. So just > > blindly adding the checksum parameter to the commandline would not > > maintain the generic support. On the other hand, I haven't found any > > other users of the script than ones where it's called with opkg as the > > type. > > ipkg-push is pretty much unmaintained on my side. I don't use it at all. > This is the only reason why it still supports ipkg. We removed the ipkg > support pretty much everywhere else a long time ago. > > I don't know much about this but it seem to me, that always having SHA256 > checksums wont hurt, right?
I think the main issue would be that the target opkg package needs to be configured to have SHA256 support, and that thus not all targets might support it? I can easily keep md5 as the default, as long as there's a clean way to switch to sha256. > > With all that, I think: > > > Suggestions: > > - Make ipkg-push just support opkg, and allow the checksumtype to be > > specified as a parameter to the script (image_ipkg.make already > > generates the local using opkg-make-index only) > > This is the way to go. Just remove the 'type' option. And, unless there is > a good reason against it, just add the '--checksum SHA256' arguments > unconditionally. > Probably for the other user of opkg-make-index as well. Sounds good, I've prepared a patch - allows the checksum type to be set in platformconfig, then propagates it to the script (replacing the 'type' argument with a 'checksum' argument instead), and also uses the selected type for the local index generation. _______________________________________________ ptxdist mailing list ptxdist@pengutronix.de
