Looking for input on "locally patch tcpdump or merge in latest release from upstream?"

[Patrick Kelsey]

Yes, I'll give it a go.  I haven't yet done a dry-run to see how hairy it
might be, so no predictions there.


Thanks,
Patrick

On Thu, May 28, 2015 at 1:18 PM, hiren panchasara <
hiren at strugglingcoder.info> wrote:

> On 05/28/15 at 10:12P, Xin Li wrote:
> > -----BEGIN PGP SIGNED MESSAGE-----
> > Hash: SHA512
> >
> > Hi, Hiren,
> >
> > On 05/27/15 21:55, hiren panchasara wrote:
> > > On 05/28/15 at 12:40P, Patrick Kelsey wrote:
> > >> Hi,
> > >>
> > >> I've had a patch for a capsicum-related issue in tcpdump sitting
> > >> around since last September (
> > >> https://lists.freebsd.org/pipermail/freebsd-current/2014-September/05
> > 2049.html)
> > >>
> > >>
> > that is still needed and that I want finally address in the tree (the pa
> > tch
> > >> was reviewed by rwatson@ and pjd@ back then).
> > >>
> > >> This issue was patched separately in the upstream tcpdump sources
> > >> in February (
> > >> https://github.com/the-tcpdump-group/tcpdump/commit/887bf88fd058f8c0e
> > f9a5af1a95b43753e3ad2eb),
> > >>
> > >>
> > along with a refactor of the associated capsicum code, and that work has
> > >> been present in tcpdump releases since 4.7.3 (
> > >> http://www.tcpdump.org/tcpdump-changes.txt).
> > >>
> > >> The last tcpdump release imported into the FreeBSD tree was 4.6.2
> > >> ( http://svnweb.freebsd.org/base/vendor/tcpdump/).
> > >>
> > >> tcpdump release import/merges have recently resulted in some
> > >> confusion/lost local patches due to the extent of the diffs
> > >> (e.g., the thread at
> > >> https://lists.freebsd.org/pipermail/svn-src-head/2015-February/067853
> > .html).
> > >>
> > >>
> > >>
> > I see three possible ways to proceed:
> > >>
> > >> 1. Apply the minimal-local-diff patch from last September to our
> > >> local tcpdump sources.  This seems like it might contribute to a
> > >> future difficult/lossy tcpdump vendor import/merge.
> > >>
> > >> 2. Import tcpdump 4.7.3 or later to address this issue.  Are
> > >> there any reasons why this might not be desired?  I don't have a
> > >> feel for when/why past tcpdump vendor imports have been performed
> > >> or avoided.
> > >>
> > >> 3. Cherry-pick the upstream patch and apply it to our local
> > >> sources, directly addressing only this issue and avoiding future
> > >> tcpdump vendor import/merge problems related to this issue.
> > >>
> > >> I'm looking for input on the above.  If left to my own devices,
> > >> I'd go with (3).
> > >
> > > Latest upstream release is 4.7.4 and the one before that was 4.6.2
> > > which we already have in the tree. I think we should get latest
> > > instead of picking bits and pieces when possible.
> > >
> > > CCing Xin for his input as he has been doing last few imports.
> >
> > Yes, I think we should do the new import.  (Are you willing to do
> > that?  I'm doing it only because nobody else were doing it...)
>
> I think Patrick volunteered to do the import. If he can't, I'll do it.
>
> Cheers,
> Hiren
>