Re: Pika buildfarm member failure on pgcrypto/test sha2 - Mailing list pgsql-hackers

From Marko Kreen
Subject Re: Pika buildfarm member failure on pgcrypto/test sha2
Date
Msg-id e51f66da1002240911g2445394q8534eeaf7464c5fc@mail.gmail.com
Whole thread Raw
In response to Re: Pika buildfarm member failure on pgcrypto/test sha2  (Rémi Zara <remi_zara@mac.com>)
List pgsql-hackers
On 2/24/10, Rémi Zara <remi_zara@mac.com> wrote:
>  Le 24 févr. 2010 à 01:04, Marko Kreen a écrit :
>  > On 2/24/10, Rémi Zara <remi_zara@mac.com> wrote:
>  >> Pika, which has been upgraded to NetBSD/mips 5.0.2, failed twice in a row pgcrypto/test sha2 because of the
followingwarning (identical each time) : 
>  >>
>  >
>  >> Anything I should try ?
>  >
>  > Please try --without-openssl.
>
>
> Indeed, it works.
>
>  My openssl version is OpenSSL 0.9.9-dev 09 May 2008 bundled with NetBSD 5.0.2.


I managed to reproduce it on 5.0.2/x86.  The problem is their
openssl library - sha384 writes out 64 bytes instead of 48.

As both OpenSSL 0.9.8 and 1.0-beta5 are fine, it's a netbsd-only
problem.   Their OpenSSL version is 0.9.9, which seems to be a
random CVS snapshot before 1.0-beta era.  Seems they did not bother
to track fixes in openssl HEAD.

>  Should I report a bug to NetBSD (and if so which one) ?

Yes - buggy sha384.

>  Is there a workaround ?

Turn OpenSSL off...

--
marko


pgsql-hackers by date:

Previous
From: "Kevin Grittner"
Date:
Subject: Re: A thought on Index Organized Tables
Next
From: Rémi Zara
Date:
Subject: Re: NaN/Inf fix for ECPG