Matthias-Christian Ott <ott%mirix.org@localhost> writes: > As mentioned in the bug report, Quagga will sent bogus IPv6 RAs if the > interface name is longer than 6 characters on NetBSD and will break your > network. Thus the bug has some urgency. > > I provided a patch that was approved by the a Quagga maintainer. > Unfortunately Greg Troxel didn't include the patch yet (I contacted him > personally and he was assigned in the bug tracker). Is there anything > that can be done about the patch? Can someone else integrate it? I didn't get to dealing with this. I have a few problems with your PR, though: There's no actual patch to pkgsrc (adding a patch file with a comment, bumping PKGREVISION, dealing with configure changes properly), with a statement that the actual patch to pkgsrc has been tested to result in the right thing happening (vs hand-built quagga with this patch being better). Upstream does not appear to have accepted the change. While the basic issue and approach seem ok, there are a lot of details to avoid other trouble. Generally I feel that upstream problems should be addressed upstream. The patch changes the zclient ABI, and this has not yet been accepted upstream. There's no bounds checking on the memcpy (kernel_socket.c:469) that the length copied fits in the target. The upstream PR has two patches, with two ids, and it's not 100% clear from the NetBSD which one you want applied. Sorry to sound so negative, but there's quite a bit of work between what's in gnats/bugzilla and what's committable.
Attachment:
pgpAjZsExICoC.pgp
Description: PGP signature