On Thu, Oct 22, 2015 at 3:38 PM, Dave S <snidely.too_at_gmail.com> wrote:
> On Thu, Oct 22, 2015 at 1:36 PM, Daniel Stenberg <daniel_at_haxx.se> wrote:
>> On Thu, 22 Oct 2015, Dave S wrote:
>>
>>> I get a sigsegv in curl_easy_perform() for certain URLs. I'm trying to
>>> access a local server on port 7070, and as I tweak the URL to try get it
>>> right, some of the combinations give me a coredump (which abrt_server takes
>>> away, not trusting me). System is a Centos 7 setup. What should I be doing
>>> to avoid the coredump?
>>
>>
>> Can you tell us more details on exactly what URLs you use to cause the
>> crashes? It is not meant to ever crash given valid input.
>
> I'm accessing a server running on another machine on the local network
> (if it matters, both are VMs, but with different hosts).
> The server is running Centos 6.6. IIRC, leaving the port (":7070" out
> the URL caused a crash, presumably because the default HTTP port was
> refusing a connenction.
>
> Currently, I am using
>
> https://10.3.171.27:7070/requests
>
> which reaches the server, but it 400's the request. The server's logs show
>
> 2015-10-22 14:09:55,339 WARN badMessage: 400 Illegal character for
> HttpChannelOverHttp_at_46f1278a{r=0,a=IDLE,uri=-}
>
> in the application log, and
>

This next entry was for the Windows client.

> 10.3.171.45 - - [22/Oct/2015:20:43:03 +0000] "POST /request
> HTTP/1.1" 400 - 107
>
> in the access log.
> (dot-45 is the system with the client, where the segfault occurs).

No, wrong client. But using http gets the libcurl client into the log
with 401 errors.

>
> Note that http://10.3.171.27:7070/requests does not segfault when it
> gets a 401 error.
>
> (but the 401 doesn't the return status non-zero)

As I said in another post, turning on FAILONERROR segfaults.

Thanks

Dave
/dps
-------------------------------------------------------------------
List admin: http://cool.haxx.se/list/listinfo/curl-library
Etiquette: http://curl.haxx.se/mail/etiquette.html
Received on 2015-10-23