On 7/7/2016 8:43 AM, Paul Jakma wrote:
> On Thu, 7 Jul 2016, Paul Jakma wrote:
>
>> Unfortunately, that means the ff, nits and pushback tracking heads are going
>> to get non-fast-forward updated.
> I've fixed the issue and updated {pushback,nits}/{a,b} and ff.
Looks like this fixes the exit problem. Two issues I see immediately are:
1 valgrind:
==12640== 192 bytes in 3 blocks are definitely lost in loss record 1 of 1
==12640== at 0x4A057BB: calloc (vg_replace_malloc.c:593)
==12640== by 0x4C27B1: zcalloc (memory.c:89)
==12640== by 0x47EF2E: bnc_new (bgp_nexthop.c:80)
==12640== by 0x4A3697: bgp_find_or_add_nexthop (bgp_nht.c:126)
==12640== by 0x43850C: bgp_start (bgp_fsm.c:722)
==12640== by 0x439528: bgp_event (bgp_fsm.c:1177)
==12640== by 0x4370DE: bgp_start_timer (bgp_fsm.c:227)
==12640== by 0x4BFC4E: thread_call (thread.c:1358)
==12640== by 0x423007: main (bgp_main.c:489)
==12640==
{
<insert_a_suppression_name_here>
Memcheck:Leak
fun:calloc
fun:zcalloc
fun:bnc_new
fun:bgp_find_or_add_nexthop
fun:bgp_start
fun:bgp_event
fun:bgp_start_timer
fun:thread_call
fun:main
}
2 some new log messages generated every seconds
2016/07/07 09:02:35 BGP: There is already write fd [14]
2016/07/07 09:02:35 BGP: There is already write fd [16]
2016/07/07 09:02:35 BGP: There is already write fd [17]
Updates of the VPN & Encap SAFIs are still not working in an RR config.
There also seems something wrong with bringing up adjacencies in
general, but again I'm out of time on this for a bit...
> Unfortunately, the new refs are not fast-forward-mergable. In the longer
> run, if volatile refs are an issue then maybe we just have to pass the
> bare commit IDs around for staging trees.
ahh, why not just different branches, e.g.,
patch-tracking/8/proposed/{v1,v2,v3}/...
They are cheap and easy? vN can contain all the changes, V(N=1) is
created when rebases are done to clean up the history/log
> Other alternatives would be not have staging trees, but then you're
> going to have very messy histories with edits and little fixes scattered
> all around the place, and intermingled with numerous reverts.
The above addresses this without forced pushes - although I'm not sure I
care so much about a messy log
Lou
> The former
> will make life harder for anyone who wants to backport fixes to an older
> release; both may make life _more_ miserable for anyone following along
> with out-of-tree work (e.g. WIP stuff, whatever).
>
> regards,
_______________________________________________
Quagga-dev mailing list
[email protected]
https://lists.quagga.net/mailman/listinfo/quagga-dev
