Re: proposal: better support for debugging of overloaded functions

Pavel Stehule <pavel.stehule@gmail.com>

From: Pavel Stehule <pavel.stehule@gmail.com>
To: Abhijit Menon-Sen <ams@toroid.org>
Cc: Robert Haas <robertmhaas@gmail.com>, pgsql-hackers@postgresql.org
Date: 2012-01-27T13:48:41Z
Lists: pgsql-hackers

Attachments

Hello

2012/1/26 Abhijit Menon-Sen <ams@toroid.org>:
> At 2011-11-24 17:44:16 +0100, pavel.stehule@gmail.com wrote:
>>
>> patch is relative long, but almost all are changes in regress tests.
>> Changes in plpgsql are 5 lines
>
> The change looks good in principle. The patch applies to HEAD with a bit
> of fuzz and builds fine… but it fails tests, because it's incomplete.
>
> Pavel, your patch doesn't contain any changes to pl_exec.c. Did you just
> forget to submit them? Anyway, some errcontext() calls need to be taught
> to print ->fn_signature rather than ->fn_name. I made those changes, and
> found some more failing tests.

It was my mistake - using fn_signature for runtime errors is good idea


>
> Updated patch attached. Ready for committer.

I found a small issue - there was uninitialized fn_signature for
online blocks so I append line

function->fn_signature = pstrdup(func_name); to
plpgsql_compile_inline(char *proc_source) function

modified patch is in attachment

Pavel

>
> -- ams