Re: gnupg for Digital UNIX
From: SL Baur <steve(a)xemacs.org>
Date: 26 May 1999 11:18:53 +0900
Sender: owner-xemacs-beta(a)xemacs.org
Kevin Oberman <oberman(a)es.net> writes in xemacs-beta(a)xemacs.org:
> The main requirement is to edit all occurrences of '-lc' out of
> configure except for the two in comment lines for checking for dlopen
> in libc.
> Once that was done, it compiled fine with:
> "setenv CC cc;setenv CFLAGS -std;configure;gmake;
> (gmake is gnu make. The default make dumps core in 'make doc'.)
> It builds cleanly, but it fails three rather critical checks,
How do you eliminate the myriad of broken preprocessor lines? The
gnupg maintainer, evidently, is unwilling to convert them to standard
C.
On DU the -std1 flag enforces ANSI compliance. -std is ANSI "with
some relaxations" and it seems to be happy with the pre-processor lines.
> encrypt.test, encryptp.test and armencryptp.test. It's my
guess that
> it is a 32 bit dependency in one of the encryption routines. I will
> continue to beat on it as time permits. These things can be painful to
> track down, but I believe I'll get there eventually.
Try eliminating the `informational' comment `Warning using insecure memory'
and see if the tests pass. With 0.9.6 and egcs I got everything but the
genkey1024 test to pass and that failure was likely due to a broken
installation of expect. I haven't tried compiling 0.9.7 with egcs yet
and I haven't come close to getting a working version with the DEC C
compiler.
Bingo. I had a feeling that the insecure message was the culprit, but
I don't see why. It looks line g10_log_info writes everything to
stderr, just like it should.
Hopefully I'll get this figured out some day.
In any case, gnupg-0.9.6 now passes all 23 tests:
gpg (GnuPG) 0.9.6
Copyright (C) 1999 Free Software Foundation, Inc.
This program comes with ABSOLUTELY NO WARRANTY.
This is free software, and you are welcome to redistribute it
under certain conditions. See the file COPYING for details.
Supported algorithms:
Cipher: 3DES, CAST5, BLOWFISH
Pubkey: ELG-E, DSA, ELG
Hash: MD5, SHA1, RIPEMD160, TIGER
PASS: version.test
PASS: mds.test
PASS: decrypt.test
PASS: decrypt-dsa.test
PASS: sigs.test
PASS: sigs-dsa.test
PASS: encrypt.test
PASS: encrypt-dsa.test
PASS: seat.test
PASS: clearsig.test
PASS: encryptp.test
PASS: detach.test
PASS: armsigs.test
PASS: armencrypt.test
PASS: armencryptp.test
PASS: signencrypt.test
PASS: signencrypt-dsa.test
PASS: armsignencrypt.test
PASS: armdetach.test
PASS: armdetachm.test
PASS: detachm.test
PASS: genkey1024.test
PASS: conventional.test
===================
All 23 tests passed
===================
I will attach the diffs for the two files I edited. The edit to secmem
is clearly a kludge and I need to find the real problem, so I should
have an additional patch before too much longer. I consider the patch
to configure to be a fix and don't have any plan to do anything else
to it.
To build (under tcsh):
setenv CC cc;setenv CFLAGS -std;configure
I still get some unresolved symbols in rndlinux and rndegd. I'm not
sure of the significance of this, but it does not cause any checks to
fail.
R. Kevin Oberman, Network Engineer
Energy Sciences Network (ESnet)
Ernest O. Lawrence Berkeley National Laboratory (Berkeley Lab)
E-mail: oberman(a)es.net Phone: +1 510 486-8634
Attachments:
- secmem.patch (text/plain — 426 bytes)
- configure.patch (text/plain — 1.1 KB)