Alan Cronin (alcronin) wrote:
> The requirement from the DoD is at:
> http://jitc.fhu.disa.mil/jitc_dri/pdfs/UCR_2013_Combined_signed.pdf -
> section 5.2.1 - requirement IP6-20. This will be switched via
> configuration in the application.
As I read them, those requirements say that nothing sh
Alan Cronin (alcronin) wrote:
> The reason for making this change is that we have some customers who are
> still using IPv4 and are trialling IPv6 but do not want any traffic going
> to those (IPv6) servers during this time. However the DNS server will
> still return A and records.
If you don
Hi all.
Here is a patch that adds a new dns resolver, async-unbound, to libcurl. This
is a first step towards the goal of full DANE support in curl.
This patch is still a bit crude and not ready for merge:
- It enables unbound by default, instead of using a configure option.
I'm not fluent in
by: Michał Górny and Anthony G. Basile.
Signed-off-by: Björn Stenberg
---
lib/connect.c | 13 ++---
1 file changed, 6 insertions(+), 7 deletions(-)
diff --git a/lib/connect.c b/lib/connect.c
index 4b6ee00..97a0655 100644
--- a/lib/connect.c
+++ b/lib/connect.c
@@ -1104,13 +1104,12 @
This fixes a rare Happy Eyeballs bug where if the first IP family runs
out of addresses before the second-family-timer fires, and the second
IP family's first connect fails immediately, no further IPs of the
second family are attempted.
---
lib/connect.c | 19 +--
1 file changed, 1
Tor Arntsen wrote:
> Re. test-patches.. something I've been thinking about for some time is
> that we could maybe have a test-branch in the Github repository where
> we could push patches still under development. I could add additional
> builds to my autobuild systems which builds from the test bra
Steve Holme wrote:
> Looking at the code the only problem I can see is in singleipconnect() when
> the default clause gets called in "switch (error)".
This indeed looks wrong, though I fail to provoke the socket leak on any of my
machines.
My attempts at getting access to a sparc machine are sti
Steve Holme wrote:
> The bad news is... The Solaris and IRIX builds are complaining about an open
> handle in connect.c:1263 when connecting to a non-listening socket:
Yeah, I saw that too. I struggle to see in the code what path brings that
error. I have asked around to see if I can locate someo
Steve Holme wrote:
> * waitconnect_getsock() then returns 1 due to the TUNNEL_CONNECT test at
> multi.c:640 but it doesn't set any of the socket handles
Yes, this is the bug. Looks like I introduced it in my original Happy Eyeballs
patch. Here is a patch to fix it.
--
Björn
From 1d2639df6d63aba
Steve Holme wrote:
> We seem to have 8 more tests that generate Valgrind errors - which all seems
> to be Proxy CONNECT related :(
Augh, I forgot to reenable valgrind in my pre-submit test! :-(
> Whilst I can fix up un-initialised structures like sockbunch in multi.c:777
> and add VALID_SOCK((soc
Guenter wrote:
> the error reported is:
> Open file descriptor created at ../../curl/lib/connect.c:1260
Here is a patch for this.
I had to remove a line from ftp.c that clears bits.tcpconnect[SECONDARYSOCKET]
at what looks to me like an odd time, but I must confess to not have a perfect
grasp o
Steve Holme wrote:
> A couple of problems in the auto builds:
>
> * The Solaris 10 build is reporting "url.c, line 5581: warning: end-of-loop
> code not reached" - while the loop breaks out because of a break statement
> it would seem that this loop is no longer required.
You are right. The purpo
Rebased on commit aa61e1 and adjusted commit message to adhere to template.
--
Björn
>From e412af6428706a6e6bb3c1368309e91db120cce9 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Bj=C3=B6rn=20Stenberg?=
Date: Tue, 29 Oct 2013 11:51:25 +0100
Subject: [PATCH] connect: Add connection delay to Happy Eyeba
Updated against commit 66b8557.
--
Björn
>From 8380ffafaa4f518814789de31468e477dc556b5d Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Bj=C3=B6rn=20Stenberg?=
Date: Tue, 29 Oct 2013 11:51:25 +0100
Subject: [PATCH] Add connection delay to Happy Eyeballs.
This patch adds a 200ms delay between the first
Steve Holme wrote:
> a) Splitting your patches up a little - this one appears to address 2 or 3
> separate issues (the add of the delay, IP address ordering, failed connect
> fix and tidy up) which would help with others reviewing them as well as
> enabling the application of individual parts if ne
Steve Holme wrote:
> I saw that most of the auto builds have mostly come back to life overnight
> and as such noticed that some tests are showing the "Whut?" connection
> issue.
>
> As such I have pushed my take on the Curl_is_connected() fix part of the
> above mentioned patch.
Actually, a separ
Paul Marks wrote:
> Note that I have not attempted to test crazy situations involving
> multiple IPv4 and IPv6 addresses, with the families interleaved.
I have tested with multiple IPv4 and IPv6 addresses for a name (such as
bad12.haxx.se) but I don't know how to make my system return them in jum
Paul Marks wrote:
> - Running on a network without IPv6:
> + Fetch ds.test-ipv6.com:
> - Success, using IPv4.
> + Fetch ipv6.test-ipv6.com:
> - Hangs for 5 minutes -- Regression!
Thank you for testing. My refactoring of the synchronous error handling code
was a smidge too aggressive.
Hi all.
This patch brings the Happy Eyeball code closer in line with the RFC6555 by
adding the requisite connection delay and address family preference.
Please help me test it.
--
Björn
>From f55341d8a948a8aa3b538d4e774b5062f76c14a8 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Bj=C3=B6rn=20Stenber
Kamil Dudka wrote:
> Maybe we should commit the change you suggest also upstream in order to
> improve the debugging experience of binary releases?
Yes, unless this attached patch fixes it. Please test it at your convenience.
--
Björn
>From 2c0328f28182cb0bd3632654f85d9cefcd877c40 Mon Sep 17 00:
Kamil Dudka wrote:
> The latest upstream git HEAD (f6c335d6) still breaks the test-suite on my
> RHEL-6 box. The tests exercising non-listening sockets (19 702 703 704 705)
> hang consuming 100% CPU and the file tests/log/trace19 consumes several GBs
> of disk space very quickly. It contains:
>
Paul Marks wrote:
> I then ran this on my router:
> $ iptables -I FORWARD -i eth1 -s 216.218.228.119 -j DROP
>
>
> And curl hangs waiting for the IPv4 response:
Here is a fix for this.
--
Björn
>From 4ea3b672ce0fe3c1c58fc0bab2b8051c5987e1bf Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Bj=C3=B6rn=2
Paul Marks wrote:
> >> As an Internet Lorax, I beg you to reconsider this algorithm.
Yes, the current algorithm is quite naive. I'm working on an update.
> I then ran this on my router:
> $ iptables -I FORWARD -i eth1 -s 216.218.228.119 -j DROP
>
>
> And curl hangs waiting for the IPv4 response
Daniel Stenberg wrote:
> Lovely. I've merged this and pushed on master now. Thanks!
Awesome!
I found a mistake. In a special case it will only try the first address in a
family. Here is the fix.
--
Björn
>From fad9d25e789398dcdb12ff1d281134d3c9fdf65d Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Bj
Daniel Stenberg wrote:
> On Sun, 27 Oct 2013, Björn Stenberg wrote:
>
> >Rebased and merged against HEAD.
>
> Thanks, but now I get a compile error:
Argh, git/brain interface malfuntion.
Here it is with the fix actually committed.
--
Björn
>From 4cf69b9fa1275d1b507058e
Rebased and merged against HEAD.
--
Björn
>From e04f4feef2a9216c8c8cbd1ea67ff6fd1929d291 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Bj=C3=B6rn=20Stenberg?=
Date: Sat, 26 Oct 2013 14:17:33 +0200
Subject: [PATCH] Add "Happy Eyeballs" for IPv4/IPv6.
This patch invokes two socket connect()s nearly si
t and thus will win.
Signed-off-by: Björn Stenberg
---
lib/connect.c | 298 -
lib/connect.h |2 -
lib/ftp.c |5 +-
lib/multi.c | 13 ++-
lib/url.c |8 +-
lib/urldata.h |2 +
6 files changed, 164 insertio
_IPRESOLVE_Vx in the libcurl api).
Björn Stenberg (1):
Add "Happy Eyeballs" for IPv4/IPv6.
lib/connect.c | 298 -
lib/connect.h |2 -
lib/ftp.c |5 +-
lib/multi.c | 13 ++-
lib/url.c |8 +-
lib/urldata.h
28 matches
Mail list logo
