[Pkg-openssl-devel] Bug#626060: Bug#626060: openssl: FTBFS on sparc64 due to configuration issue

[Kurt Roeckx]

On Sun, May 08, 2011 at 04:44:31PM +0200, Aurelien Jarno wrote:
> On Sun, May 08, 2011 at 04:13:24PM +0200, Kurt Roeckx wrote:
> > On Sun, May 08, 2011 at 03:32:39PM +0200, Aurelien Jarno wrote:
> > > Package: openssl
> > > Version: 1.0.0d-2
> > > Severity: important
> > > Tags: patch
> > > User: debian-sparc at lists.debian.org
> > > Usertags: sparc64
> > > 
> > > openssl FTBFS on sparc64 with a testsuite failure. This is due to a
> > > small configuration issue, fixed by the patch below. I changed
> > > SIXTY_FOUR_BIT_LONG into BN_LLONG, as it is done for debian-sparc-v9 or
> > > solaris64-sparc-v9. Could you please apply this patch in the next
> > > upload. Thanks in advance.
> > 
> > What is the size of a "long"?  64 bit? Then as far as I know you
> > should use SIXTY_FOUR_BIT_LONG.
> 
> Yes, the size of a long is 64 bit. If in this case SIXTY_FOUR_BIT_LONG
> should be used, do you know why solaris64-sparcv9-gcc, 
> solaris64-sparcv9-cc, BSD-sparc64 and linux64-sparcv9 use BN_LLONG? 
> I basically did the change based on them, and it fixed the testsuite
> issue.

BSD-sparc64 used to use SIXTY_FOUR_BIT_LONG, but doesn't anymore,
and I don't know why.

I have no idea why they make things so complicated with all those
defines anyway.  I think they have the same pieces of code using
"long" or "long long" as variable instead of using a typedef or
something.

> > If long is 64 bit, as it is on most of our 64 bit ports, I would
> > look into why it's failing instead.
> > 
> 
> The problem is that test BN_add. I will try to investigate why.

I have no idea why it should break other than a long is not
64 bit.


Kurt