Horton

Re: Add BufFileRead variants with short read and EOF detection

Peter Eisentraut <peter.eisentraut@enterprisedb.com>

View thread View on postgr.es
From: Peter Eisentraut <peter.eisentraut@enterprisedb.com>
To: Amit Kapila <amit.kapila16@gmail.com>
Cc: pgsql-hackers <pgsql-hackers@postgresql.org>
Date: 2023-01-12T09:14:11Z
Lists: pgsql-hackers

Attachments

On 10.01.23 07:20, Amit Kapila wrote:
>> The existing uses of %m are wrong.  This was already fixed once in
>> 7897e3bb902c557412645b82120f4d95f7474906, but the affected areas of code
>> were apparently developed at around the same time and didn't get the
>> fix.  So I have attached a separate patch to fix this first, which could
>> be backpatched.
>>
> +1. The patch is not getting applied due to a recent commit.
> 
>> The original patch is then rebased on top of that.  I have adjusted the
>> error message to include the file set name if available.
>>
>> What this doesn't keep is the purpose of the temp file in some cases,
>> like "hash-join temporary file".  We could maybe make this an additional
>> argument or an error context, but it seems cumbersome in any case.
>>
> Yeah, we can do that but not sure if it is worth doing any of those
> because there are already other places that don't use the exact
> context.

Ok, updated patches attached.

Commits

  1. Add BufFileRead variants with short read and EOF detection

  2. Fix some BufFileRead() error reporting

  3. Push lpp variable closer to usage in heapgetpage()

  4. Fix buffile.c error handling.