On 3 Jan 2003 at 21:40, Niels Möller wrote:
"Paul Swartz" z3p@twistedmatrix.com writes:
The first error seems to be from src/nettle/md5.h, with inttypes.h not found. configure detects that the file does not exist, but LSH seems to ignore it.
Do you have <stdint.h> instead? I would expect both files to be included with cygwin's gcc package, but I don't know.
I don't have either. I hacked one together that's based off <sys/types.h> however, and it seems to work pretty well, except...
Has anyone else tried this successfully, or am I going to have to wade through the errors?
As far as I know, no-one has compiled lsh on cygwin before. So you have to bug report the problems you find, and try to fix them. I don't think there should be any fundamental problem to get the client to work. The server may be harder, I don't know how login is handled on win32 or cygwin. And I guess a win32 "native" server would be more useful.
Adding the inttypes.h to /usr/include solved most of the problems, until I got to the test suite. At this point, all hell breaks loose. I get undefined reference errors to _nettle_aes[128|192|256] in aes-test.c, to _cbc_encrypt and _cbc_decrypt, _md5_init, _md5_update, _rsa_md5_sign, _rsa_md5_verify, _sha1_init, _sha1_update, _rsa_sha1_sign, _rsa_sha1_verify, _dsa_signature_init, _knuth_lfib_init, _knith_lfib_random, _dsa_sign, _dsa_verify, and _dsa_signature_clear in testutils.c.
Trying with a CVS checkout version of nettle, the linking fails on aes- text.exe, with _aes_encrypt_ and _aes_decrypt undefined.
Another bug is that nettle/desdata.c compiled twice, first without the -D flags, the second time with them. I worked around this by removing the #if HAVE_CONFIG_H around the #include.
-p