Experimental Physics and
Industrial Control System

Michael Davidsaver via Core-talk <[email protected]> · Wed, 10 Apr 2019 09:28:47 -0700

On 4/10/19 8:39 AM, Kasemir, Kay via Core-talk wrote:
> From: Bruno Martins <[email protected]>
>> Related:  https://epics.anl.gov/core-talk/2018/msg00284.php
> Alright, so missing info about size encoding is a known issue.
> 
> Also related:
> The element count encoding of the search request for the channel names,
>   struct {
>         int searchInstanceID;
>         string channelName;
>     } channels[];
> },
> always seems to use a 2-byte short count, not the usual size encoding.

yup.

https://github.com/epics-base/pvAccessCPP/blob/78410499f0236f5830a0937567bbb794de7b2347/src/server/responseHandlers.cpp#L281

https://github.com/mdavidsaver/cashark/blob/7ea746615292e33ed44d6f3a54c6181996804a33/pva.lua#L379-L383

> I assume that's because all implementations currently start with
>   short count: 0
> and then keep updating the buffer in place:

Also yes.

https://github.com/epics-base/pvAccessCPP/blob/78410499f0236f5830a0937567bbb794de7b2347/src/remote/channelSearchManager.cpp#L256-L275

the "ick!!!" comment is mine.

https://github.com/epics-base/pvAccessCPP/blob/78410499f0236f5830a0937567bbb794de7b2347/src/remote/channelSearchManager.cpp#L57-L58

>   short count: 1
>   int some_channel_id
>   string some_channel_name
> 
>   short count: 2
>   int some_channel_id
>   string some_channel_name
>   int some_channel_id
>   string some_channel_name
> 
> With normal size encoding, once you have count >= 254 you'd have to shift all channel ids and names down to make room for a larger count.
> 
> -Kay
> 

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System