Alec,

Am 28.04.19 um 14:24 schrieb Alec Liu:
> Hi List,
> 
> Got it.
> This is how to reproduce the issue I mentioned.
> 
> 1) I am using commit 4de0eba8488f7d7b471d60da76d815c69e44fecb, the
> master branch;
> 2) setup a tcp relay for HTTP as frontend and HTTPS as backend;
> 3) make sure the HTTPS connection is working, when starting the haproxy;
> 4) try to access the HTTP frontend, confirm it is working;
> 5) then reroute the ip address of the HTTPS connection to a black
> hole; (when I am using socks4 proxy for upstream will be easier to do
> it, just turn the proxy off)
> 6) try to access the HTTP frontend, then it crash.

I'm afraid I am unable to reproduce the issue. I am using the same
commit and the exact same configuration.

1. ./haproxy -d -f ./crasher.cfg
2. I open http://localhost:20936/haproxy?stats in the web browser.
3. sudo iptables -A OUTPUT --dst 169.38.103.48 -j REJECT
4. I wait a bit and reload http://localhost:20936/haproxy?stats

Am misunderstanding your instructions?

Best regards
Tim Düsterhus

> Please check the attached file for more details.
> 
> Thank you.
> 
> Regards,
> Alexander Liu
> 
> 
> On Sun, Apr 28, 2019 at 7:57 PM Tim Düsterhus <[email protected]> wrote:
>>
>> Alec,
>>
>> Am 28.04.19 um 09:51 schrieb Alec Liu:
>>> When I am testing my patch which is base on the master branch, and
>>> come across some crash causing by segmentation fault.
>>>
>>> And I have it retested without my patch applied, confirmed nothing to
>>> do with my patch.
>>> I have a quick test with the git tag v1.9.0 as well, looks like
>>> working normally.
>>>
>>> I am wondering if I should report it or just leave it, since it is a
>>> development branch issue.
>>
>> Yes, please do report it! In the worst case the issue is already known,
>> in the best case it is not and you found a legitimate issue that would
>> not have been found otherwise. So nothing to lose here.
>>
>> For a segmentation fault a full backtrace using gdb or valgrind would be
>> helpful. The other things (config, haproxy -vv, etc. all apply as well).
>>
>> Best regards
>> Tim Düsterhus

Reply via email to