Re: BUG #8970: ts_parse incorrectly split numbers in digit token

On 25/01/2014 16:43, Bruce Momjian wrote:
> On Sat, Jan 25, 2014 at 10:39:53AM -0500, Bruce Momjian wrote:
>> Uh, there are some known problems with Cygwin, particularly with text
>> search. I don't think anyone knows the cause, but I think it is problem
>> in the Cygwin platform.
>>
>> Here is a thread that mentions the issue:
>>
>> http://www.postgresql.org/message-id/51B59794.3000500@gmail.com
>>
>> and here is a comment about the failures:
>>
>> http://www.postgresql.org/message-id/52E1EBD0.4090304@dunslane.net
>>
>> AFAICT the regression is in Cygwin. The buildfarm passes because it's
>> using an oldish Cygwin release, 1.7.7 rather than the current 1.7.27. I
>> have brought the regression the athe attention of the Cygwin people in
>> the past, but without response.

Bruce,
I took over, the binary package responsabilities, in Jan 2013 with 9.2.2
http://cygwin.com/ml/cygwin-announce/2013-01/msg00032.html

Before that the previous package releasse was 8.2.11
http://cygwin.com/ml/cygwin-announce/2009-01/msg00010.html

Can we move forward ?
Testing on 1.7.7 will not help me on packaging a full functional binary
deployed through our setup as our users expect.

> Oops, I just replied to you with your own thread. :-) I thought it was
> odd we had all these Cygwin reports all of a sudden. LOL

Always me ;-)

> Anyway, are you saying these are not platform bugs as Andrew suggested?

what do you you mean as platform bugs ?

I have no clue of how "ts_parse" works, so from my point of view
it could be any of:

1) wrong expectation of postgres about cygwin behaviour
(I noticed that the code is full of "#if defined(platform)"
that could break if the platform evolves)
2) a bug in latest cygwin library (cygwin core or newlib )
3) a bug in one of the several other library that postgres is using
4) a bug in the latest version of C compiler
In the past, on other softwares, I have seen any of these happening.

So where I should look in the postgres code to understand why "123"
is splitted in 3 token ?

If (2) or (3) I need to pass a "simple" test case to the other
core and package maintainers that shows the incorrect behaviour
and allow to amend the bug or report to the upstream library developers.

If (1) we need to amend postgres code, and for (4)
report a test case to gcc developers and look for a temporary workaround

Regards
Marco