Bug#1029944: neon27 FTBFS on IPV6-only buildds
Hi,
On Sat, Feb 11, 2023 at 7:48 PM Andreas Henriksson wrote:
> If replacing "127.0.0.1" with "localhost" does not work the attached
> (completely untested) patch might be an option instead (simply skip the
> test on EAFNOSUPPORT).
I don't have access to an IPv6 host to test on, so I just think it
would not work.
> (Sorry if the patch is completely broken, but I hope you get the
> idea...)
This is broken for sure as some tests create a server socket (get the
'Address family for hostname not supported' error) and then next tests
would like to connect to it ('request failed: Could not connect to
proxy server: Connection refused' error) or simply get an other error
('request failed with 5 not NE_ERROR' message). While your patch would
skip the former, the latter would still be a problem.
I can remove the IPv4 only tests that are detected on buildds, but
there's at least one which hangs ('Build killed with signal TERM after
150 minutes of inactivity').
Any IPv6 porterbox available?
Regards,
Laszlo/GCS
Bug#1029944: neon27 FTBFS on IPV6-only buildds
Hello again,
On Sat, Feb 11, 2023 at 06:55:13PM +0100, Andreas Henriksson wrote:
> On Sun, Jan 29, 2023 at 02:29:23PM +0100, László Böszörményi (GCS) wrote:
> > Control: tags -1 +confirmed
> >
> > On Sun, Jan 29, 2023 at 12:33 PM Adrian Bunk wrote:
> > > https://buildd.debian.org/status/logs.php?pkg=neon27=amd64
> > >
> > > ...
> > > auth.. 3/20 FAIL - retries (line 311: HTTP error:
> > > Could not resolve hostname `127.0.0.1': Address family for hostname not
> > > supported)
> > Is there a way to detect such buildds as a maintainer? What can I do
> > except notifying upstream and / or disable such tests?
If replacing "127.0.0.1" with "localhost" does not work the attached
(completely untested) patch might be an option instead (simply skip the
test on EAFNOSUPPORT).
(Sorry if the patch is completely broken, but I hope you get the
idea...)
Regards,
Andreas Henriksson
--- test/utils.h.orig 2023-02-11 19:38:52.122713689 +0100
+++ test/utils.h 2023-02-11 19:40:25.601514834 +0100
@@ -25,7 +25,7 @@
#include "child.h"
-#define ONREQ(x) do { int _ret = (x); if (_ret) { t_context("line %d: HTTP error:\n%s", __LINE__, ne_get_error(sess)); return FAIL; } } while (0);
+#define ONREQ(x) do { int _ret = (x); if (_ret) { t_context("line %d: HTTP error:\n%s", __LINE__, ne_get_error(sess)); if (strcmp(ne_get_error(sess), "Could not resolve hostname `127.0.0.1': Address family for hostname not supported") == 0) { return SKIP; } else { return FAIL; } } } while (0);
int single_serve_string(ne_socket *s, void *userdata);
Bug#1029944: neon27 FTBFS on IPV6-only buildds
On Sun, Jan 29, 2023 at 02:29:23PM +0100, László Böszörményi (GCS) wrote: > Control: tags -1 +confirmed > > On Sun, Jan 29, 2023 at 12:33 PM Adrian Bunk wrote: > > https://buildd.debian.org/status/logs.php?pkg=neon27=amd64 > > > > ... > > auth.. 3/20 FAIL - retries (line 311: HTTP error: > > Could not resolve hostname `127.0.0.1': Address family for hostname not > > supported) > Is there a way to detect such buildds as a maintainer? What can I do > except notifying upstream and / or disable such tests? FWIW "127.0.0.1" is hardcoded here: https://sources.debian.org/src/neon27/0.32.5-1/test/utils.c/#L207 Possibly that could be replaced with "localhost" instead, but I'm not sure if using 127.0.0.1 is an attempt at hiding that the code is ipv4-only rather than protocol independent (maybe)... I tried quickly looking at upstream git history if it could confirm my suspicion, but the closest I could find was: https://github.com/notroj/neon/commit/eff6521be537c74511593743bf8665d898e26567 Seems like localhost -> 127.0.0.1 switch was intentional (and done during SVN days?). Maybe someone digging deeper can find "r1910". Regards, Andreas Henriksson
Bug#1029944: neon27 FTBFS on IPV6-only buildds
On Sun, Jan 29, 2023 at 02:29:23PM +0100, László Böszörményi (GCS) wrote: > Control: tags -1 +confirmed > > On Sun, Jan 29, 2023 at 12:33 PM Adrian Bunk wrote: > > https://buildd.debian.org/status/logs.php?pkg=neon27=amd64 > > > > ... > > auth.. 3/20 FAIL - retries (line 311: HTTP error: > > Could not resolve hostname `127.0.0.1': Address family for hostname not > > supported) > Is there a way to detect such buildds as a maintainer? What can I do > except notifying upstream and / or disable such tests? I don't know any better solution than fixing or disabling such tests (but I am not an expert on that). > Cheers, > Laszlo/GCS cu Adrian
Bug#1029944: neon27 FTBFS on IPV6-only buildds
Control: tags -1 +confirmed On Sun, Jan 29, 2023 at 12:33 PM Adrian Bunk wrote: > https://buildd.debian.org/status/logs.php?pkg=neon27=amd64 > > ... > auth.. 3/20 FAIL - retries (line 311: HTTP error: > Could not resolve hostname `127.0.0.1': Address family for hostname not > supported) Is there a way to detect such buildds as a maintainer? What can I do except notifying upstream and / or disable such tests? Cheers, Laszlo/GCS
Bug#1029944: neon27 FTBFS on IPV6-only buildds
Source: neon27 Version: 0.32.1-1 Severity: serious Tags: ftbfs https://buildd.debian.org/status/logs.php?pkg=neon27=amd64 ... auth.. 3/20 FAIL - retries (line 311: HTTP error: Could not resolve hostname `127.0.0.1': Address family for hostname not supported) auth.. 3/20 auth.. 4/20 auth.. 4/20 auth.. 4/20 auth.. 5/20 E: Build killed with signal TERM after 150 minutes of inactivity
