Darcy Buskermolen wrote:
> On Thursday 13 July 2006 13:15, Christopher Browne wrote:
>   
>> Darcy Buskermolen wrote:
>>     
>>> On Tuesday 11 July 2006 11:38, Darcy Buskermolen wrote:
>>>       
>>>> On Tuesday 11 July 2006 09:43, Christopher Browne wrote:
>>>>         
>>>>> I have set up tarballs for a first release candidate; Jan should be
>>>>> uploading those soon, and I'll see about updating the web site when
>>>>> they are uploaded...
>>>>>           
>>>> Umm re-wrap RC-.  code in CVS would not compile on older versions of
>>>> GCC. (you can't declare vars in the middle of a control block).  I have
>>>> since fixed this.
>>>>         
>>> Ok there is more breakage under olderversions of PG.  For example
>>> continue is not a valid loop destructor in plpgsql for PG versions prior
>>> to 8.1 We will need to refactor slon_quote_input to not use these
>>> features.  I'll look at this a bit later but my available time to work on
>>> this is somewhat limited.  Any chance you can look at this Jan since you
>>> were the last to alter this function.
>>>       
>> I'm looking at it; the CONTINUE looks to be relatively unproblematic to
>> get rid of as it's not an escape from a deep control structure.
>>
>> There's still something with a couple of exceptions; probably that we
>> need to have variables for them rather than computing things dynamically  
>> in the RAISE request...
>>     
>
> See the latest email from  Stefan Kaltenbrunner more more issues with the 
> function.
>   
There was also something wrong with version handling in slonik.c...

I have a nagging suspicion that I broke versioning when I was turning
off support for 7.3.

That's revised much more properly now.

Further, if you try to connect Slonik to an unsuitable database, it will
give substantially better error messages now.

- If you hook up to a 7.2 DB, you'll get "(versions < 7.3 were never
supported by Slony-I)"

- If you hook up to a 7.3 DB, you'll get "(try Slony-I 1.1.5)"

I have checked in that fix.

I have checked in a fix that unrolls the CONTINUE.

It sounds like "standards_conforming_strings" is something we need to
talk about further.  Seems like something where our timing is very right
to fix it now, and thus avoid forcing 8.2 users to do any "instant
upgrades."  (In some ways, that's a lot to hope for, since 8.2 doesn't
really exist, and probably won't be released 'til ~November, but we
might as well avoid any anvils that we see hanging over our heads :-).)