boost::asio::read_until Segmentation Fault & Third Party Signal Handler
My Linux application sis using bool::asio::read_until() to read data from a socket. The application also uses commercial, 3rd party hardware. Occasionally my program crashes due to a SIGSEGV violation. The core dump stack trace is shown below. It indicates that the 3rd party signal handler is invoked while my thread is waiting for socket I/O, and that the crash is occurring during signal handler execution. While I recognize that this is an issue with the 3rd party vendor (and I am currently engaged in discussions with them), I'm wondering if my usage of boost is incorrect, or if there is something that I should be doing differently in performing socket I/O? I am using boost v1.43. [cid:image001.png@01D17303.98B2DD10] E. Thomas Craig Principal Software Engineer SWIFT Division KLA-Tencor Corporation 408-875-7909 (work) 650-799-4787 (cell)
On Mon, Feb 29, 2016 at 4:16 PM, Craig, Tom
I am using boost v1.43.
Is it possible for you to retest using a more recent version of Boost? v1.43 is six years old now. I am not the asio maintainer, nor even very familiar with it. But when I am charged with debugging a problem in my own software, my first question is: does it still reproduce with the current production version? If not, the facile answer is to direct the user to update.
Hi Tom - I hope you are well. On 2/29/16 13:16, Craig, Tom wrote:
My Linux application sis using bool::asio::read_until() to read data from a socket. The application also uses commercial, 3^rd party hardware. Occasionally my program crashes due to a SIGSEGV violation.
The core dump stack trace is shown below. It indicates that the 3^rd party signal handler is invoked while my thread is waiting for socket I/O, and that the crash is occurring during signal handler execution.
While I recognize that this is an issue with the 3^rd party vendor (and I am currently engaged in discussions with them), I’m wondering if my usage of boost is incorrect, or if there is something that I should be doing differently in performing socket I/O?
It is a little hard to know from the dump if the code is correct; however, starting in Boost 1.47 release you can easily setup a signal handler. http://www.boost.org/doc/libs/1_47_0/doc/html/boost_asio/overview/signals.ht...
I am using boost v1.43.
<snip picture> Michael Caisse Ciere Consulting
Michael, Thanks. All is well (except for the segmentation violation we are experiencing). I hope all is well with you. I don't want to set up a signal handler. I am using 3rd party libraries, and they are setting up their own signal handler (likely not using boost - we don't control this). The segmentation violation occurs when their signal handler is activated. It would seem that their libraries are either not creating the signal handler properly, or are doing so in a way that is not compatible with boost. Any suggestions that you (or others) have about how to address this problem are appreciated. We are, of course, pursuing this with the vendor that wrote the signal handler. [cid:image001.png@01D17303.98B2DD10] E. Thomas Craig Principal Software Engineer SWIFT Division KLA-Tencor Corporation 408-875-7909 (work) 650-799-4787 (cell) From: Boost-users [mailto:boost-users-bounces@lists.boost.org] On Behalf Of Michael Caisse Sent: Tuesday, March 01, 2016 1:12 PM To: boost-users@lists.boost.org Subject: Re: [Boost-users] boost::asio::read_until Segmentation Fault & Third Party Signal Handler Hi Tom - I hope you are well. On 2/29/16 13:16, Craig, Tom wrote: My Linux application sis using bool::asio::read_until() to read data from a socket. The application also uses commercial, 3rd party hardware. Occasionally my program crashes due to a SIGSEGV violation. The core dump stack trace is shown below. It indicates that the 3rd party signal handler is invoked while my thread is waiting for socket I/O, and that the crash is occurring during signal handler execution. While I recognize that this is an issue with the 3rd party vendor (and I am currently engaged in discussions with them), I'm wondering if my usage of boost is incorrect, or if there is something that I should be doing differently in performing socket I/O? It is a little hard to know from the dump if the code is correct; however, starting in Boost 1.47 release you can easily setup a signal handler. http://www.boost.org/doc/libs/1_47_0/doc/html/boost_asio/overview/signals.ht...http://www.boost.org/doc/libs/1_47_0/doc/html/boost_asio/overview/signals.ht... I am using boost v1.43. <snip picture> Michael Caisse Ciere Consulting
participants (3)
-
Craig, Tom -
Michael Caisse -
Nat Goodspeed