There are a few changes in glibc-2.10 that might make your software fail to build with it, please refer to my blog post [1] if you're not sure what the problem is. And no I don't usually provide emerge --info with these bugs because they are caused by glibc-2.10! Thanks, Diego [1] http://blog.flameeyes.eu/2009/05/24/c-libraries-galore
Created attachment 194511 [details] Build log
Intresting. I didn't manage to get so far with this one, sockets.c:492: Error: Incorrect register `%edx' used with `q' suffix sockets.c:470: Error: Incorrect register `%edx' used with `q' suffix sockets.c:2159: Error: Incorrect register `%eax' used with `q' suffix sockets.c:2160: Error: Incorrect register `%eax' used with `q' suffix sockets.c:1746: Error: Incorrect register `%eax' used with `q' suffix sockets.c:1748: Error: Incorrect register `%eax' used with `q' suffix sockets.c:2099: Error: Incorrect register `%eax' used with `q' suffix sockets.c:2040: Error: Incorrect register `%eax' used with `q' suffix
This is fixed in ircservices-5.1.20 (just released).
(The getline error, that is. The sockets.c errors are all in FD_SET and FD_CLR macro calls, so that looks like a glibc bug.)
Fixed in portage
(In reply to comment #3) > This is fixed in ircservices-5.1.20 (just released). > Please open a new bug for the version bump (if you want)
(In reply to comment #4) > (The getline error, that is. The sockets.c errors are all in FD_SET and FD_CLR > macro calls, so that looks like a glibc bug.) These sockets.c errors are actually caused by: defs.h /* Make sizeof() return an int regardless of compiler (avoids printf * argument type warnings). */ #define sizeof(v) ((int)sizeof(v)) You've just got to live with sizeof() returning size_t like it should. Removing this line fixes those asm compilation errors on amd64 systems. Just FYI ;-).