Re: Closing some 8.4 open items - Mailing list pgsql-hackers

From Magnus Hagander
Subject Re: Closing some 8.4 open items
Date
Msg-id 49DCEB2A.6050901@hagander.net
Whole thread Raw
In response to Re: Closing some 8.4 open items  (Tom Lane <tgl@sss.pgh.pa.us>)
List pgsql-hackers
Tom Lane wrote:
> Magnus Hagander <magnus@hagander.net> writes:
>> Heikki Linnakangas wrote:
>>> cygwin supports POSIX_FADV_SEQUENTIAL (and POSIX_FADV_NORMAL to revert
>>> it), but not any of the other flags. It maps it to
>>> NtSetInformationFile() like this:
> 
>> We set this in our open() wrapper in the code today.
> 
> Really?  Where?  I didn't find any of the mentioned symbols in a quick
> grep.

We pass FILE_FLAG_SEQUENTIAL_SCAN to the open call if O_SEQUENTIAL is
specified.


> I'm not sure how Windows interprets FILE_SEQUENTIAL_ONLY, but if that
> really means that it assumes *only* sequential accesses will happen,
> I'm not sure that we'd want to turn it on.

It's an access-optimization hint, that's all.

//Magnus


pgsql-hackers by date:

Previous
From: Tom Lane
Date:
Subject: Re: Closing some 8.4 open items
Next
From: Dave Page
Date:
Subject: Re: Closing some 8.4 open items