[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]
Re: security/gnupg on AIX
-----BEGIN PGP SIGNED MESSAGE-----
On Feb 25, 2009, at 12:58 PM, Louis Guillaume wrote:
To get security/gnupg to build on AIX I had to do 2 things. The
first was easy, but the second I'm not comfortable with (although
things do work).
1. Added "CONFIGURE_ARGS+= --disable-asm" to the Makefile
2. Compiled cipher/rijndael.c manually with `-O1'.
To do the second thing, I did "bmake configure" then...
# cc -DHAVE_CONFIG_H -I. -I.. -I.. -I../include -I../intl -
DGNUPG_LIBDIR="\"/usr/pkg/lib/gnupg\"" -I/usr/pkg/include -I/usr/
include -D_THREAD_SAFE -q64 -qarch=ppc64 -O1 -qstrict -qcpluscmt -
qkeyword=inline -ma -I/usr/pkg/include -I/usr/include -c rijndael.c
cc: 1501-210 (W) command option -O contains an incorrect subargument
... the warning doesn't seem to have any adverse effect at runtime
(for me anyway).
The second thing didn't actually prevent the package from building,
but at runtime I had errors like this when decrypting...
Rijndael-128 test encryption failed.
gpg: key setup failed: selftest failed
gpg: decryption failed: selftest failed
... compiling rijndael.c with `-O1' fixes this. I don't really know
why, but I found the solution by Googling. People were blaming
compilers; so I think it may be an xlc issue??
Please advise if there's a way to get this patched in pkgsrc. Thanks!
Did you use -O >2 ? I dimly remember xlc docs saying that higher
optimization levels enable 'potentially semantics changing'
optimization strategies - maybe you just found such a case. Not sure
if that was at -O2 or higher though, and the xlc in question was
ancient - 5.0.2 or something like that.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (Darwin)
-----END PGP SIGNATURE-----
Main Index |
Thread Index |