Re: [boost] [stacktrace] Two days remaining before review ends

Subject: Re: [boost] [stacktrace] Two days remaining before review ends
From: Antony Polukhin (antoshkka_at_[hidden])
Date: 2017-04-27 06:15:12

2017-04-27 8:55 GMT+03:00 Gottlob Frege via Boost <boost_at_[hidden]>:
> On Sat, Apr 22, 2017 at 5:41 AM, Antony Polukhin via Boost
> <boost_at_[hidden]> wrote:
>> 2017-04-11 4:20 GMT+03:00 Emil Dotchevski via Boost <boost_at_[hidden]>:
>>> I have a question on stacktrace: is it possible to decouple the capture of
>>> stacktraces from the conversion to a human-readable format? These are two
>>> different use cases and while most programs need both, that is not true for
>>> libraries.
>>
>> Done. Now if you include boost/stacktrace/stacktrace.hpp you would be
>> able to catch stacktraces but would not be able to decode it.
>> Now boost/stacktrace/stacktrace.hpp does not include windows.h, COM,
>> LexicalCast and other "heavy" Boost libraries
>>
>
> Similarly, what I'd like is to grab a stacktrace, which is basically
> just addresses,
> then later match that up with, for example, pdb files, to get function
> names etc.
>
> When I push as task onto an async queue, I'd like to capture the
> stacktrace of the code doing the push,
> and include that with the task in the queue.
> So when the thread that performs the task throws or crashes, I can log
> "who" pushed that task.
> Then, back at work, decode that log and say "here, it's your fault".
>
> But the capture needs to be fast and small.  Thus just addresses.

Yes. That's how the library works. Captures and stores only addresses,
decoding is on demand.

-- 
Best regards,
Antony Polukhin