Niels Möller wrote: >Ewald Wasscher <[EMAIL PROTECTED]> writes: > >>After compiling liboop and libgmp (I disabled zlib) lsh 1.3.6 seems to >>work fine. >> Erm, not completely. For some reason password authentication is broken when compiling with uClibc and running on the development host. When compiling with glibc it works fine. I suppose it's a problem with uClibc and md5-shadow passwords. Am I right that there's no "UseLogin" parameter to lshd as a workaround?
>>IIRC I had to edit the makefiles for liboop so that it >>wouldn't be linked with libnsl, and because of that glibc's >>libc.so.6........ >> > >Good to hear that. I think Dan Egnor (liboop author) would change >liboop's configure script the same way I hacked lsh's, if you just >explain to him why it is necessary. > I think I'll try. > > >We've exchanged a few mails, but as I'm neither using nor >understanding uClibc, I'm probably not the right person to do that. > >My interpretation of the matter is that the conflict > actually libnsl depends on libc.so.6 >between libnsl >and glibc (and perhaps the mere existence of libnsl) is a bug in the >uClibc environment. > Could be, my interpretation is that the configure script finds headers and libraries from the host- (and not the cross-compile-) environment. > > >BTW, the configure script hack, which replaces AC_CHECK_LIB with >AC_SEARCH_LIB, seems to be useful also on Cray's Unicos OS, as that >linker seems to accept any -lfoo option, and just prints annoying >warnings about the library not existing. I have *almost* got >lsh-1.2.5 to compile on Lysator's old Cray YMP running unicos 9.0.2. > Cool :-) Greetings from the Netherlands, Ewald Wasscher
