Welcome! Log In Create A New Profile

Advanced

haproxy tests with sanitizers: bug found ? https://gitlab.com/chipitsine/haproxy/-/jobs/80667862

Posted by Илья Шипицин 
Hello,

I had an idea, what if we will run reg-tests with various sanitizers ?

for example,

https://gitlab.com/chipitsine/haproxy/-/jobs/80667862

(gcc + sanitize-address).
can someone confirm a bug there ?

"AddressSanitizer: heap-use-after-free on"


I'll add more sanitizers later.
thanks for your attention!

Ilya Shipitsin
just for the record, asan was very easy to enable:

https://gitlab.com/chipitsine/haproxy/blob/master/.gitlab-ci.yml

вт, 10 июл. 2018 г. в 16:35, Илья Шипицин <[email protected]>:

> Hello,
>
> I had an idea, what if we will run reg-tests with various sanitizers ?
>
> for example,
>
> https://gitlab.com/chipitsine/haproxy/-/jobs/80667862
>
> (gcc + sanitize-address).
> can someone confirm a bug there ?
>
> "AddressSanitizer: heap-use-after-free on"
>
>
> I'll add more sanitizers later.
> thanks for your attention!
>
> Ilya Shipitsin
>
On 07/10/2018 01:35 PM, Илья Шипицин wrote:
> Hello,

Hello Ilya,

> I had an idea, what if we will run reg-tests with various sanitizers ?
>
> for example,
>
> https://gitlab.com/chipitsine/haproxy/-/jobs/80667862
>
> (gcc + sanitize-address).
> can someone confirm a bug there ?

I confirm there is a little bug, but without any consequence (only a
memleak). In fact the address sanitizer detects that the loggin buffers
are allocated twice by the main thread.

I am going to provide a patch which makes your test pass in another thread.

Thank you.
On 07/11/2018 10:41 AM, Frederic Lecaille wrote:
> On 07/10/2018 01:35 PM, Илья Шипицин wrote:
>> Hello,
>
> Hello Ilya,
>
>> I had an idea, what if we will run reg-tests with various sanitizers ?
>>
>> for example,
>>
>> https://gitlab.com/chipitsine/haproxy/-/jobs/80667862
>>
>> (gcc + sanitize-address).
>> can someone confirm a bug there ?
>
> I confirm there is a little bug, but without any consequence (only a
> memleak). In fact the address sanitizer detects that the loggin buffers
> are allocated twice by the main thread.

So in fact there is no bug (see Christopher answer:
https://www.mail-archive.com/[email protected]/msg30622.html).

After googling, it seems that libasan may produce false positives.

You must not use it to run reg tests.

Fred.
Thanks, I'll check it out.

ср, 11 июл. 2018 г. в 18:20, Frederic Lecaille <[email protected]>:

> On 07/11/2018 10:41 AM, Frederic Lecaille wrote:
> > On 07/10/2018 01:35 PM, Илья Шипицин wrote:
> >> Hello,
> >
> > Hello Ilya,
> >
> >> I had an idea, what if we will run reg-tests with various sanitizers ?
> >>
> >> for example,
> >>
> >> https://gitlab.com/chipitsine/haproxy/-/jobs/80667862
> >>
> >> (gcc + sanitize-address).
> >> can someone confirm a bug there ?
> >
> > I confirm there is a little bug, but without any consequence (only a
> > memleak). In fact the address sanitizer detects that the loggin buffers
> > are allocated twice by the main thread.
>
> So in fact there is no bug (see Christopher answer:
> https://www.mail-archive.com/[email protected]/msg30622.html).
>
> After googling, it seems that libasan may produce false positives.
>
> You must not use it to run reg tests.
>
> Fred.
>
Sorry, only registered users may post in this forum.

Click here to login