The lambdamoo ebuild I authored was recently accepted into the tree. One of the tweaks to the ebuild seems to have broken crypt() such that it uses cleartext instead of actually encrypting the string. This is disastrous for passwords, to say the least. Specifically, it's the "autoconf" in src_unpack() that is the culprit. It compiles fine for me with it commented out, and crypt() then works as expected. What is the rationale behind having the "autoconf" there? Additionally, I've added a patch that enables outgoing networks conenctions so that new players can be emailed their password. This requires care in the database, though, to prevent abuse. I will attach a diff for the ebuild as well as the patch file it should apply.
Created attachment 14400 [details, diff] removes "autoconf" to fix crypt() behavior and adds step to enable outbound connections
Created attachment 14401 [details] patches the source to enable outbound connections. Belongs in files/
By the way, the bug for the original submission of the ebuild is 14312.
the reason for the autoconf is because the configure script was generated by a very *old* version of autoconf ... ive never seen this thing happen before, but my fault for not checking it out more added your changes to cvs, 1.8.1-r1 now exists
I just tried this replacement configure.in, and it seems to work fine. I ran autoconf on it and built. crypt() works as expected. It's old, but not quite as old as the one in the tarball. Would it be better to drop this in and run autoconf before compiling? http://sourceforge.net/tracker/index.php?func=detail&aid=429134&group_id=3692&atid=303692
the package installs to the correct locations right now so i'd say lets not worry about it ;) the packages works off of 1 binary (/usr/bin) and 1 database file (we tell lambdamoo where it is anyways) so unless it breaks on someone's box lets just leave as is, thanks for the link though :)