configure.ac has: AC_CHECK_FUNCS([pam_putenv]) yet auth-pam.c uses pam_putenv regardless: if (sshpam_rhost != NULL) { debug("PAM: setting PAM_RHOST to \"%s\"", sshpam_rhost); sshpam_err = pam_set_item(sshpam_handle, PAM_RHOST, sshpam_rhost); if (sshpam_err != PAM_SUCCESS) { pam_end(sshpam_handle, sshpam_err); sshpam_handle = NULL; return (-1); } /* Put SSH_CONNECTION in the PAM environment too */ pam_putenv(sshpam_handle, sshpam_conninfo); } This is despite earlier uses of pam_putenv being wrapped with #ifdef HAVE_PAM_PUTENV: #ifdef HAVE_PAM_PUTENV /* Errors are not fatal here */ if ((r = pam_putenv(sshpam_handle, env)) != PAM_SUCCESS) { error("PAM: pam_putenv: %s", pam_strerror(sshpam_handle, r)); } #endif
Created attachment 3281 [details] Possible patch
Created attachment 3282 [details] Add no-op implementation of pam_putenv I think we would be better off adding a no-op implementation of pam_putenv similar to what we already do with pam_getenvlist. Your diff has some other unrelated changes in configure.ac, what's the deal with those?
(In reply to Darren Tucker from comment #2) > Created attachment 3282 [details] > Add no-op implementation of pam_putenv > > I think we would be better off adding a no-op implementation of > pam_putenv similar to what we already do with pam_getenvlist. > > Your diff has some other unrelated changes in configure.ac, what's > the deal with those? Oops, my bad. Internal changes. Let me test your patch on our HP-UX systems.
Thanks for the report. Patch (attachment #3282 [details]) has been committed and will be in the next release. I'm pretty sure that will resolve it, but if you discover any problems please reopen.
closing resolved bugs as of 8.6p1 release