Re: try/catch in scsynth

On 2018-03-29 15:25, [email protected] wrote:
> If there's no try/catch, then failure to allocate memory in any
> plug-in means the server would crash abruptly, without any sort of
> informative message to the user. It's not really optional here. (It's
> not necessary for plug-in code to throw exceptions intentionally for
> try/catch to be meaningful. And plug-ins I think should pretty
> obviously avoid using exceptions for routine flow-of-control.)
> 
> I guess, if try/catch were bad for performance in the absence of any
> exceptions, we would have heard many many complaints by now. And, if
> you do have exceptions (e.g. out of memory), you have bigger problems
> than an audio dropout.

i'm aware of what this try/catch is protecting from -- my question was 
whether it's the safest way to handle this. i think you're right that 
there isn't a performance problem with try/catch unless something goes 
wrong, in which case the slight risk of an xrun is the least of your 
worries.

scott carver explained to me on slack that there is a more subtle 
problem here. i think that we might be throwing C++ exceptions across 
plain C binary boundaries (twice), which is undefined behavior. i am not 
sure in this case because the throw and try/catch are both in scsynth, 
but it's happening through the plugin shared library. see tim's comment 
in 
https://github.com/supercollider/supercollider/pull/2483#discussion_r87257083


nathan

_______________________________________________
sc-dev mailing list

info (subscription, etc.): http://www.birmingham.ac.uk/facilities/ea-studios/research/supercollider/mailinglist.aspx
archive: http://www.listarc.bham.ac.uk/marchives/sc-dev/
search: http://www.listarc.bham.ac.uk/lists/sc-dev/search/