Experimental Physics and
Industrial Control System

Caution: This email originated from outside of Cosylab.


On 7/20/22 11:51, Jure Varlec wrote:
>  > imo. it is reasonable to join worker threads here, but I would
>>recommend against deleting.
>>
>>Please keep in mind that delete leaves dangling pointers in the
>>asyn port.  imo. this is an invitation for the chaos of
>>use-after-free errors.
>
> Could be, but I can't see how.

Ah, some Famous Last Words :)


> The IOC is shutting down, asynManager is rejecting queueRequest calls to a disabled port

Which could be re-enabled.  And assuming that all code
actually respects the enabled switch.


> (which is what the device support code uses), and asyn doesn't register any hooks that I can see.

Currently.


> So AFAICT, when the records are the only users of the port, there should be no chaos.

As far as has been said, no.  This is an assumption through.

Actually, I should have mentioned before now the "atExitDebug"
global, which logs each exit hook before it runs.  I added this
after once being surprised by an exit hook I didn't know about.

> var atExitDebug 1


> Are you just trying to be careful,

In short, yes.

I'm trying to be careful and encourage what I see a good practice.
I certainly don't intend to jump on you in particular.

Tech-talk is an archived public mailing frequented by new(er)
users.  So in addition to Érico, I'm conscious of the audience.
Both now, and those whom google will later bring.


> or are you aware of something I missed? If I'm wrong about this, I'd love to know before I have to face the chaos :) .
>
> Best,
> Jure