Re: reg-tests are broken when running osx + openssl

[Frederic Lecaille]

On 5/6/19 11:25 AM, Frederic Lecaille wrote:

On 5/6/19 9:54 AM, Илья Шипицин wrote:

there's some random failure

https://travis-ci.com/haproxy/haproxy/jobs/197824840

looks like test is not stable


According to the log, S13 and S37 syslog servers have the same port: 42464


Obviously this is a vtest bug. I could sometimes reproduce it. I cannot 
reproduce anymore with this commit: : 
https://github.com/haproxyFred/VTest/commit/17871365a14d81234426ac203da3af30bb46e506


I have sent a PR for vtest for this issue: 
https://github.com/vtest/VTest/pull/13.


Fred.

Re: reg-tests are broken when running osx + openssl

[Frederic Lecaille]

On 5/6/19 9:54 AM, Илья Шипицин wrote:

there's some random failure

https://travis-ci.com/haproxy/haproxy/jobs/197824840

looks like test is not stable


According to the log, S13 and S37 syslog servers have the same port: 42464


Fred.

Re: reg-tests are broken when running osx + openssl

[Илья Шипицин]

surprisingly, LibreSSL builds are green

пн, 6 мая 2019 г. в 13:03, Willy Tarreau :

> On Mon, May 06, 2019 at 12:54:06PM +0500,  ??? wrote:
> > there's some random failure
> >
> > https://travis-ci.com/haproxy/haproxy/jobs/197824840
> >
> > looks like test is not stable
>
> isn't it the same at the other one you reported that looked related to
> the latest SSL fixes ?
>
> Willy
>

Re: reg-tests are broken when running osx + openssl

[Willy Tarreau]

On Mon, May 06, 2019 at 12:54:06PM +0500,  ??? wrote:
> there's some random failure
> 
> https://travis-ci.com/haproxy/haproxy/jobs/197824840
> 
> looks like test is not stable

isn't it the same at the other one you reported that looked related to
the latest SSL fixes ?

Willy

Re: reg-tests are broken when running osx + openssl

[Илья Шипицин]

there's some random failure

https://travis-ci.com/haproxy/haproxy/jobs/197824840

looks like test is not stable

пн, 6 мая 2019 г. в 11:11, Willy Tarreau :

> On Sun, May 05, 2019 at 10:11:38PM +0200, Frederic Lecaille wrote:
> > I had no access to any travis environment when it has been told in
> previous
> > mails that /tmp could not work, and /var/tmp could not either.
> > They were the first tested values.
> >
> > Now that I have setup a travis account, I have just double checked, and
> > contrary to what has been previously told, /tmp value for TMPDIR may make
> > the reg tests work on OS X with travis.
> >
> > So here is attached to this mail a better fix.
>
> Excellent, thank you Fred for testing and for the explanation! Patch
> merged now, we'll see if things improve.
>
> Willy
>

Re: reg-tests are broken when running osx + openssl

[Willy Tarreau]

On Sun, May 05, 2019 at 10:11:38PM +0200, Frederic Lecaille wrote:
> I had no access to any travis environment when it has been told in previous
> mails that /tmp could not work, and /var/tmp could not either.
> They were the first tested values.
> 
> Now that I have setup a travis account, I have just double checked, and
> contrary to what has been previously told, /tmp value for TMPDIR may make
> the reg tests work on OS X with travis.
> 
> So here is attached to this mail a better fix.

Excellent, thank you Fred for testing and for the explanation! Patch
merged now, we'll see if things improve.

Willy

Re: reg-tests are broken when running osx + openssl

[Frederic Lecaille]

On 5/5/19 9:51 AM, Willy Tarreau wrote:

On Fri, May 03, 2019 at 07:27:48PM +0200, Frederic Lecaille wrote:

-  - env VTEST_PROGRAM=../vtest/vtest make reg-tests
+  - env TMPDIR=~/tmp VTEST_PROGRAM=../vtest/vtest make reg-tests


That may sound like a stupid question, but what makes you think that
~/tmp will actually result in a smaller path ? I mean ~ might very
well be "/home/automated-builds/travis/version-1.2.3.4/haproxy" on
the target system. Thus I think we're trying random locations until
it works :-/


I had no access to any travis environment when it has been told in 
previous mails that /tmp could not work, and /var/tmp could not either.

They were the first tested values.

Now that I have setup a travis account, I have just double checked, and 
contrary to what has been previously told, /tmp value for TMPDIR may 
make the reg tests work on OS X with travis.


So here is attached to this mail a better fix.


Fred.
>From dd2828071bc36375a485a7ab7d6d601d9ce729be Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Fr=C3=A9d=C3=A9ric=20L=C3=A9caille?= 
Date: Fri, 3 May 2019 19:16:02 +0200
Subject: [PATCH] BUILD: travis: TMPDIR replacement.
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

TMPDIR default value may be too long to create UNIX sockets for the stats
used during the reg tests. Indeed vtest builds its temporary working directory
${tmpdir} variable from TMPDIR variable, with /tmp as value if not already set.
This is the case on Linux contrary to OS X which sets TMPDIR with a too much long
value.

With this path we revert the part of 88c63a6 commit which tried to shorten this
TMPDIR value modifying script/run-regtests.sh. Unfortunately this was not
sufficient. Furthermore this patch force TMPDIR to /tmp value for all the OS'es.

Thank you to Tim Düsterhus and Ilya for having helped on this issue.
---
 .travis.yml | 8 +---
 1 file changed, 1 insertion(+), 7 deletions(-)

diff --git a/.travis.yml b/.travis.yml
index f689fe982..be3207113 100644
--- a/.travis.yml
+++ b/.travis.yml
@@ -27,16 +27,10 @@ install:
   # Special flags due to: https://github.com/vtest/VTest/issues/12
   - make -C ../vtest FLAGS="-O2 -s -Wall"
 
-before_script:
-  # This is a fix for the super long TMPDIR on Mac making
-  # the unix socket path names exceed the maximum allowed
-  # length.
-  - sed -i'.original' '/TESTDIR=.*haregtests/s/haregtests-.*XX/regtest.XXX/' scripts/run-regtests.sh
-
 script:
   - make CC=$CC V=1 TARGET=$TARGET $FLAGS
   - ./haproxy -vv
-  - env VTEST_PROGRAM=../vtest/vtest make reg-tests
+  - env TMPDIR=/tmp VTEST_PROGRAM=../vtest/vtest make reg-tests
 
 after_failure:
   - |
-- 
2.11.0

Re: reg-tests are broken when running osx + openssl

[Willy Tarreau]

On Fri, May 03, 2019 at 07:27:48PM +0200, Frederic Lecaille wrote:
> -  - env VTEST_PROGRAM=../vtest/vtest make reg-tests
> +  - env TMPDIR=~/tmp VTEST_PROGRAM=../vtest/vtest make reg-tests

That may sound like a stupid question, but what makes you think that
~/tmp will actually result in a smaller path ? I mean ~ might very
well be "/home/automated-builds/travis/version-1.2.3.4/haproxy" on
the target system. Thus I think we're trying random locations until
it works :-/

Theorically /tmp is always shorter. Now if /tmp doesn't work for
whatever reason we may end up having no other choice, but if so it
was not clear in this thread and further is not mentioned in the
commit message, which leaves me with a big doubt.

Willy

Re: reg-tests are broken when running osx + openssl

[Илья Шипицин]

I would even use shorter path, i.e.

mkdir ~/t

пт, 3 мая 2019 г. в 22:27, Frederic Lecaille :

> On 5/3/19 5:35 PM, Frederic Lecaille wrote:
> > On 5/3/19 3:44 PM, Илья Шипицин wrote:
> >>
> >>
> >> пт, 3 мая 2019 г. в 18:42, Tim Düsterhus  >> >:
> >>
> >> Ilya,
> >>
> >> Am 03.05.19 um 15:39 schrieb Илья Шипицин:
> >>  > when I played with enabling travis-ci, I tried to set TMPDIR
> >> directly,
> >>  > however I was not lucky enough.
> >>  > Later Tim added "sed" magic to .travis.yml
> >>  >
> >>  > personally, I do not understand why "sed" is better than
> >> assigning TMPDIR
> >>  > directly.
> >>
> >> I did not try using TMPDIR=/tmp or something like that, because I
> >> thought there must be a reason why it's that strange long path.
> >>
> >>
> >> I tried /tmp and /var/tmp
> >> it seems that not any filesystem on osx can hold network socket (at
> >> least from my point of view)
> >
> > try to create a working directory owned by the user which run the reg
> > test :
> >
> > $ mkdir -p ~/tmp/
> > $ TMPDIR=~/tmp make reg-tests
>
> I confirm that with such a value everything work on all OS'es
> (https://travis-ci.com/haproxyFred/haproxy)
>
> The attached patch should fix this issue.
>
> Thank you Tim, Ilya.
>
> Fred.
>

Re: reg-tests are broken when running osx + openssl

[Frederic Lecaille]

On 5/3/19 5:35 PM, Frederic Lecaille wrote:

On 5/3/19 3:44 PM, Илья Шипицин wrote:



пт, 3 мая 2019 г. в 18:42, Tim Düsterhus >:


    Ilya,

    Am 03.05.19 um 15:39 schrieb Илья Шипицин:
 > when I played with enabling travis-ci, I tried to set TMPDIR
    directly,
 > however I was not lucky enough.
 > Later Tim added "sed" magic to .travis.yml
 >
 > personally, I do not understand why "sed" is better than
    assigning TMPDIR
 > directly.

    I did not try using TMPDIR=/tmp or something like that, because I
    thought there must be a reason why it's that strange long path.


I tried /tmp and /var/tmp
it seems that not any filesystem on osx can hold network socket (at 
least from my point of view)


try to create a working directory owned by the user which run the reg 
test :


    $ mkdir -p ~/tmp/
    $ TMPDIR=~/tmp make reg-tests


I confirm that with such a value everything work on all OS'es 
(https://travis-ci.com/haproxyFred/haproxy)


The attached patch should fix this issue.

Thank you Tim, Ilya.

Fred.
>From fc9decae9ec679038dc494ad612dd3eb144de408 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Fr=C3=A9d=C3=A9ric=20L=C3=A9caille?= 
Date: Fri, 3 May 2019 19:16:02 +0200
Subject: [PATCH] BUILD: travis: TMPDIR replacement.
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

TMPDIR default value is too long especially on OSX systems.
We decided to shorten it for all the OS'es.

Thank you to Tim Düsterhus and Ilya for having helped on this issue.
---
 .travis.yml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/.travis.yml b/.travis.yml
index f689fe982..7475ad028 100644
--- a/.travis.yml
+++ b/.travis.yml
@@ -31,12 +31,12 @@ before_script:
   # This is a fix for the super long TMPDIR on Mac making
   # the unix socket path names exceed the maximum allowed
   # length.
-  - sed -i'.original' '/TESTDIR=.*haregtests/s/haregtests-.*XX/regtest.XXX/' scripts/run-regtests.sh
+  - mkdir ~/tmp
 
 script:
   - make CC=$CC V=1 TARGET=$TARGET $FLAGS
   - ./haproxy -vv
-  - env VTEST_PROGRAM=../vtest/vtest make reg-tests
+  - env TMPDIR=~/tmp VTEST_PROGRAM=../vtest/vtest make reg-tests
 
 after_failure:
   - |
-- 
2.11.0

Re: reg-tests are broken when running osx + openssl

[Frederic Lecaille]

On 5/3/19 3:44 PM, Илья Шипицин wrote:



пт, 3 мая 2019 г. в 18:42, Tim Düsterhus >:


Ilya,

Am 03.05.19 um 15:39 schrieb Илья Шипицин:
 > when I played with enabling travis-ci, I tried to set TMPDIR
directly,
 > however I was not lucky enough.
 > Later Tim added "sed" magic to .travis.yml
 >
 > personally, I do not understand why "sed" is better than
assigning TMPDIR
 > directly.

I did not try using TMPDIR=/tmp or something like that, because I
thought there must be a reason why it's that strange long path.


I tried /tmp and /var/tmp
it seems that not any filesystem on osx can hold network socket (at 
least from my point of view)


try to create a working directory owned by the user which run the reg test :

   $ mkdir -p ~/tmp/
   $ TMPDIR=~/tmp make reg-tests

Re: reg-tests are broken when running osx + openssl

[Илья Шипицин]

пт, 3 мая 2019 г. в 18:42, Tim Düsterhus :

> Ilya,
>
> Am 03.05.19 um 15:39 schrieb Илья Шипицин:
> > when I played with enabling travis-ci, I tried to set TMPDIR directly,
> > however I was not lucky enough.
> > Later Tim added "sed" magic to .travis.yml
> >
> > personally, I do not understand why "sed" is better than assigning TMPDIR
> > directly.
>
> I did not try using TMPDIR=/tmp or something like that, because I
> thought there must be a reason why it's that strange long path.
>

I tried /tmp and /var/tmp
it seems that not any filesystem on osx can hold network socket (at least
from my point of view)


>
> Best regards
> Tim Düsterhus
>

Re: reg-tests are broken when running osx + openssl

[Tim Düsterhus]

Ilya,

Am 03.05.19 um 15:39 schrieb Илья Шипицин:
> when I played with enabling travis-ci, I tried to set TMPDIR directly,
> however I was not lucky enough.
> Later Tim added "sed" magic to .travis.yml
> 
> personally, I do not understand why "sed" is better than assigning TMPDIR
> directly.

I did not try using TMPDIR=/tmp or something like that, because I
thought there must be a reason why it's that strange long path.

Best regards
Tim Düsterhus

Re: reg-tests are broken when running osx + openssl

[Илья Шипицин]

пт, 3 мая 2019 г. в 18:33, Frederic Lecaille :

> On 5/3/19 1:34 PM, Tim Düsterhus wrote:
> > Fred,
> > Ilya,
>
> Hello Tim,
>
> > Am 03.05.19 um 13:20 schrieb Frederic Lecaille:
> >> About the test which fail, I would say that such errors are not
> >> negligible :
> >>
> >>  Starting frontend GLOBAL: cannot change UNIX socket ownership
> >> [/var/folders/nz/vv4_9tw56nv9k3tkvyszvwg8gn/T//regtest.zHu/
> >
> > I believe this is an issue with the long TMPDIR that I tried to mitigate
> > with this:
> https://github.com/haproxy/haproxy/blob/master/.travis.yml#L34
>
> With your patch, vtest is able to create the LOG files at the same place
> $TMPDIR/ where the UNIX stats socket should be
> created. So this does not interfere with the test.
>
> > While debugging I noticed that the validation did not properly account
> > for the temporary extension of the filename during start-up, causing
> > HAProxy to accept the filename during the check, but fail to set it up.
> > This leads to the misleading error message.
>
> Yes, perhaps he UNIX stats socket filename is too long (I have found 104
> max length for sun_path on Max OS X, 108 on Linux).
>
> So, I propose you revert your fix, and try to find another ways to set
> TMPDIR with a shorter value than the default one which is too long for
> UNIX sockets. At least this is the correct way to change the working
> directory for vtest.
>
> For instance we have:
>
>
> /var/folders/nz/vv4_9tw56nv9k3tkvyszvwg8gn/T//regtest.zHu/vtc.23058.0fa4d8bc/h1/stats.sock
>
> which is 94 bytes long. Should work only if we do not add an ..tmp
> extension bigger than 10 bytes. I guess this is not the case when the
> PID is big. Now I understand why some test may pass.
>
> I have also noted that there is a missing closing bracket in this log line:
>
> ***  h10.0 debug|[ALERT] 122/093540 (23139) : Starting frontend
> GLOBAL: cannot change UNIX socket ownership
> [/var/folders/nz/vv4_9tw56nv9k3tkvyszvwg8gn/T//regtest.zHu/
>
> which is built like that:
>
>  snprintf(errmsg, errlen, "%s [%s]", msg, path);
>
> with 100 as errlen value: "cannot change UNIX socket ownership
> [/var/folders/nz/vv4_9tw56nv9k3tkvyszvwg8gn/T//regtest.zHu/" is
> exactly a 100 bytes long string. So here the path for the UNIX socket is
> truncated in the log.
>
> So let's try with a shorter TMPDIR variable please. This should fix the
> issue.
>

when I played with enabling travis-ci, I tried to set TMPDIR directly,
however I was not lucky enough.
Later Tim added "sed" magic to .travis.yml

personally, I do not understand why "sed" is better than assigning TMPDIR
directly.

please enable travis-ci.com on your accounts and try your ideas (with osx).


>
> > I did not get around to investigating this further and filing a bug
> > report, however.
> >
> > Best regards
> > Tim Düsterhus
> >
>
>

Re: reg-tests are broken when running osx + openssl

[Frederic Lecaille]

On 5/3/19 1:34 PM, Tim Düsterhus wrote:

Fred,
Ilya,


Hello Tim,


Am 03.05.19 um 13:20 schrieb Frederic Lecaille:

About the test which fail, I would say that such errors are not
negligible :

     Starting frontend GLOBAL: cannot change UNIX socket ownership
[/var/folders/nz/vv4_9tw56nv9k3tkvyszvwg8gn/T//regtest.zHu/


I believe this is an issue with the long TMPDIR that I tried to mitigate
with this: https://github.com/haproxy/haproxy/blob/master/.travis.yml#L34


With your patch, vtest is able to create the LOG files at the same place 
$TMPDIR/ where the UNIX stats socket should be 
created. So this does not interfere with the test.



While debugging I noticed that the validation did not properly account
for the temporary extension of the filename during start-up, causing
HAProxy to accept the filename during the check, but fail to set it up.
This leads to the misleading error message.


Yes, perhaps he UNIX stats socket filename is too long (I have found 104 
max length for sun_path on Max OS X, 108 on Linux).


So, I propose you revert your fix, and try to find another ways to set 
TMPDIR with a shorter value than the default one which is too long for 
UNIX sockets. At least this is the correct way to change the working 
directory for vtest.


For instance we have:

/var/folders/nz/vv4_9tw56nv9k3tkvyszvwg8gn/T//regtest.zHu/vtc.23058.0fa4d8bc/h1/stats.sock

which is 94 bytes long. Should work only if we do not add an ..tmp 
extension bigger than 10 bytes. I guess this is not the case when the 
PID is big. Now I understand why some test may pass.


I have also noted that there is a missing closing bracket in this log line:

***  h10.0 debug|[ALERT] 122/093540 (23139) : Starting frontend 
GLOBAL: cannot change UNIX socket ownership 
[/var/folders/nz/vv4_9tw56nv9k3tkvyszvwg8gn/T//regtest.zHu/


which is built like that:

snprintf(errmsg, errlen, "%s [%s]", msg, path);

with 100 as errlen value: "cannot change UNIX socket ownership 
[/var/folders/nz/vv4_9tw56nv9k3tkvyszvwg8gn/T//regtest.zHu/" is 
exactly a 100 bytes long string. So here the path for the UNIX socket is 
truncated in the log.


So let's try with a shorter TMPDIR variable please. This should fix the 
issue.



I did not get around to investigating this further and filing a bug
report, however.

Best regards
Tim Düsterhus

Re: reg-tests are broken when running osx + openssl

[Tim Düsterhus]

Fred,
Ilya,

Am 03.05.19 um 13:20 schrieb Frederic Lecaille:
> About the test which fail, I would say that such errors are not
> negligible :
> 
>     Starting frontend GLOBAL: cannot change UNIX socket ownership
> [/var/folders/nz/vv4_9tw56nv9k3tkvyszvwg8gn/T//regtest.zHu/

I believe this is an issue with the long TMPDIR that I tried to mitigate
with this: https://github.com/haproxy/haproxy/blob/master/.travis.yml#L34

While debugging I noticed that the validation did not properly account
for the temporary extension of the filename during start-up, causing
HAProxy to accept the filename during the check, but fail to set it up.
This leads to the misleading error message.

I did not get around to investigating this further and filing a bug
report, however.

Best regards
Tim Düsterhus

Re: reg-tests are broken when running osx + openssl

[Frederic Lecaille]

On 5/3/19 1:20 PM, Frederic Lecaille wrote:

So on OSX you should try to use/create a temporary working director 
where you have enough permissions to create a stats UNIX socket with 
0600 as permissions.


I meant you should try to create a temporary working directory for vtest 
using the TMPDIR environment variable, as follows for instance:


  $ mkdir ~/tmp/foo

  $ TMPDIR=~/tmp/foo make reg-tests 
reg-tests/http-capture/multiple_headers.vtc


## Preparing to run tests ##
Testing with haproxy version: 2.0-dev2-a48237-261
Target : linux2628
Options : +EPOLL -KQUEUE -MY_EPOLL -MY_SPLICE +NETFILTER +PCRE -PCRE_JIT 
-PCRE2 -PCRE2_JIT +POLL -PRIVATE_CACHE +THREAD -PTHREAD_PSHARED -REGPARM 
-STATIC_PCRE -STATIC_PCRE2 +TPROXY +LINUX_TPROXY +LINUX_SPLICE +LIBCRYPT 
+CRYPT_H -VSYSCALL -GETADDRINFO +OPENSSL +LUA +FUTEX +ACCEPT4 
-MY_ACCEPT4 +ZLIB -SLZ +CPU_AFFINITY -TFO -NS +DL +RT -DEVICEATLAS 
-51DEGREES -WURFL -SYSTEMD -OBSOLETE_LINKER +PRCTL

## Gathering tests to run ##
  Add test: reg-tests/http-capture/multiple_headers.vtc
## Starting vtest ##
Testing with haproxy version: 2.0-dev2-a48237-261
#top  TEST reg-tests/http-capture/multiple_headers.vtc TIMED OUT 
(kill -9)
#top  TEST reg-tests/http-capture/multiple_headers.vtc FAILED 
(10.010) signal=9

1 tests failed, 0 tests skipped, 0 tests passed
## Gathering results ##
## Test case: reg-tests/http-capture/multiple_headers.vtc ##
## test results in: 
"/home/flecaille/tmp/foo/haregtests-2019-05-03_13-24-59.2P0ECZ/vtc.1327.1fe5daa1"

 c15.0 HTTP rx timeout (fd:9 5000 ms)
Makefile:971: recipe for target 'reg-tests' failed
make: *** [reg-tests] Error 1


As you can see the logs are now in /home/flecaille/tmp/foo (with ~ my 
home directory: /home/flecaille)


The LOG file is here: 
/home/flecaille/tmp/foo/haregtests-2019-05-03_13-24-59.2P0ECZ/vtc.1327.1fe5daa1/LOG


and the UNIX stats socket is here:

$ grep stats 
/home/flecaille/tmp/foo/haregtests-2019-05-03_13-24-59.2P0ECZ/vtc.1327.1fe5daa1/LOG 

 h 0.0 conf|\tstats socket 
"/home/flecaille/tmp/foo/haregtests-2019-05-03_13-24-59.2P0ECZ/vtc.1327.1fe5daa1/h/stats.sock" 
level admin mode 600

Re: reg-tests are broken when running osx + openssl

[Frederic Lecaille]

On 5/3/19 11:39 AM, Илья Шипицин wrote:

Hello,

I'm expanding openssl matrix.
here's failing build

https://travis-ci.org/chipitsine/haproxy-1/jobs/527683332


Hello Ilya,

In fact this has nothing to see with openssl. A lot of tests without any 
usage of TLS/SSL also fail.


There are a lot of HTTP rx which timed out.

Only these two tests passed:

   reg-tests/http-capture/multiple_headers.vtc
   reg-tests/spoe/wrong_init.vtc

but in these cases we do not have any log.

About the test which fail, I would say that such errors are not negligible :

Starting frontend GLOBAL: cannot change UNIX socket ownership 
[/var/folders/nz/vv4_9tw56nv9k3tkvyszvwg8gn/T//regtest.zHu/


I have simulated it with such a patch on Linux:

$ git diff src/proto_uxst.c
diff --git a/src/proto_uxst.c b/src/proto_uxst.c
index 980a22649..b5f945b9f 100644
--- a/src/proto_uxst.c
+++ b/src/proto_uxst.c
@@ -309,6 +309,10 @@ static int uxst_bind_listener(struct listener 
*listener, char *errmsg, int errle

goto err_unlink_temp;
}

+   err |= ERR_FATAL | ERR_ALERT;
+   msg = "cannot change UNIX socket ownership";
+   goto err_unlink_temp;
+
ready = 0;
ready_len = sizeof(ready);
if (getsockopt(fd, SOL_SOCKET, SO_ACCEPTCONN, , 
_len) == -1)



I got the same results as yours: lots of HTTP RX time out because 
haproxy exited unespectedly.


But in such a case on my PC only reg-tests/spoe/wrong_init.vtc succeeds.
I do not understand how reg-tests/http-capture/multiple_headers.vtc can 
succeed on your side.


Would be interesting to run it on OSX with this command:

$ make reg-tests reg-tests/http-capture/multiple_headers.vtc -- --debug


So on OSX you should try to use/create a temporary working director 
where you have enough permissions to create a stats UNIX socket with 
0600 as permissions.


And let's see if that fixes your issue.


Fred.

reg-tests are broken when running osx + openssl

[Илья Шипицин]

Hello,

I'm expanding openssl matrix.
here's failing build

https://travis-ci.org/chipitsine/haproxy-1/jobs/527683332