[racket] SIGSEGV MAPERR si_code 1 fault on addr 0xb7600000

I had restarted the program 5 minutes after it crashed and it's been
running for 20+ hours handling hundreds more emails without the fault.

This is very intermittent. Which of course is good and bad. Bad for
debugging. It might be hours or days until it faults again.

So in parallel to enabling core dumps and so on, I think I'll want to
have some other process check every to see if it's still running. If
not, launch it again.

On Fri, Mar 25, 2011 at 7:18 PM, Greg Hendershott
<greghendershott at gmail.com> wrote:
> Hi, Neil. Thank you for the quick reply and good things to consider.
>
> This is 32-bit.
>
> I'd posted a week or two ago that Racket was looking for older
> versions of libssl and libcrypto than what the Amazon Linux supplies.
> I'd solved that (I thought?) by making symlinks. Sounds like I should
> revisit that.
>
> On Fri, Mar 25, 2011 at 6:36 PM, Neil Van Dyke <neil at neilvandyke.org> wrote:
>> Greg Hendershott wrote at 03/25/2011 05:52 PM:
>>>
>>> SIGSEGV MAPERR si_code 1 fault on addr 0xb7600000
>>
>> A few thoughts:
>>
>> * That address looks special or suspicious to me.  Problem with a shared
>> library, perhaps FFI?
>>
>> * If this VM is 64-bit, are you running 64-bit code and libraries?
>>
>> * You can run under "gdb" and try to get it to fail again, to get a
>> backtrace .  If you don't want to run under "gdb", enable core dumps for the
>> process, try to reproduce the failure, and then you can use "gdb" to get a
>> post-mortem backtrace from the "core" file.
>>
>> * If the backtrace doesn't help, I suggest looking through the application
>> code and any libraries you're using, for things that seem like they might be
>> problematic.
>>
>> * I don't know how you're doing SMTP, but if you have various MTAs
>> connecting with various SMTP accents, perhaps (this is wild speculation) the
>> problem occurs when one MTA tries to to SSL/TLS and there's some shared
>> library problem with OpenSSL or other library?
>>
>> * Good luck, and don't be scared off too easily.  FWIW, I can tell you that
>> clients of mine have numerous production installations of Racket-based
>> Internet apps (mostly PLT version 4.x; not much 5.x yet), including some use
>> of EC2, and I don't recall encountering a problem like this.
>>
>> --
>> http://www.neilvandyke.org/
>>
>