Re: [HACKERS] Explicit subtransactions for PL/Tcl

On Thu, 9 Mar 2017 11:12:09 +0100
Pavel Stehule <pavel.stehule@gmail.com> wrote:


> the regress tests is unstable
> 
> the proc name has mutable pid
> 
> !     (procedure "__PLTcl_proc_16503" line 3)
>       invoked from within
> ! "__PLTcl_proc_16503 SPI"
> 
> Regards

Really, I don't know what can be done with it, short of rewriting all
tests as tap-tests.

Definitely this patch is not the right place for reversing desing
decision of PL/Tcl authors to add a numeric suffix to the proc names.
(it is not PID. It is OID of the function).

Of course, I can catch all the errors inside Tcl code and return
just message, but it would sufficiently narrow test functionality.

Now test demonstrate how errors uncaught on the Tcl level interact with
postgresql error system.


                With best regards, Victor

--                Victor Wagner <vitus@wagner.pp.ru>