This version fixes some bugs, also it is the first version that builds/works with openssl-1.x and thus it should go stable together with openssl-1. Add arch teams here whenever appropriate.
My compilation fails with: i686-pc-linux-gnu-gcc -I/usr/local/ssl/include -O3 -march=athlon-xp -fomit-frame-pointer -pipe -Wall -O3 -march=athlon-xp -fomit-frame- pointer -pipe -Wall -Wl,-O1 -Wl,--as-needed \ sha_drv.c \ \ -L/usr/local/ssl/lib -lssl -lcrypto \ -DHAVE_SSL \ -L/usr/lib/erlang/lib/erl_interface-3.6.5/lib -lerl_interface -lei \ -I/usr/lib/erlang/lib/erl_interface-3.6.5/include -I/usr/lib/erlang/usr/include \ -o ../sha_drv.so \ -fpic -shared sha_drv.c:23:25: error: openssl/md2.h: No such file or directory sha_drv.c: In function ‘sha_drv_control’: sha_drv.c:40: error: ‘MD2_DIGEST_LENGTH’ undeclared (first use in this function) sha_drv.c:40: error: (Each undeclared identifier is reported only once sha_drv.c:40: error: for each function it appears in.) sha_drv.c:42: warning: implicit declaration of function ‘MD2’ make[1]: *** [../sha_drv.so] Error 1 make[1]: Leaving directory `/tmp/more/portage/portage/net-im/ejabberd-2.1.5/work/ejabberd-2.1.5/src/tls' make: *** [all-recursive] Error 1 Further up in build.log regarding ssl it shows: looking for openssl in /usr/local/ssl... checking for SSL_new in -lssl... yes checking openssl/ssl.h usability... yes checking openssl/ssl.h presence... yes checking for openssl/ssl.h... yes openssl found in /usr/local/ssl But I don't even have the directory /usr/local/ssl !? Ideas?
Dennis, please create new ticket, attach full build log there and add me in CC and a blocker on this bug. Thank you!
Created attachment 242249 [details, diff] openssl-1.0.0a mdc2 patch
It turns out that my dev-libs/openssl-1.0.0a doesn't have <openssl/md2.h> but rather <openssl/mdc2.h>. I'm not sure why MD2 was renamed to MDC2 :S. Anywho, I've attached a patch, for src/tls/sha_drv.c, (and added the epatch statement in src_prepare()), although this probably should be done upstream?
Hmm -- it appears that mdc2 is not the same thing as md2 :P. It also appears as though they fixed the md2 issue (by autodetecting whether it exists) upstream already, with their next version. :)
Dennis, I've asked you to do two things: 1.) open new bug report and 2.) attach build.log. How do you want me to help if you ignore my requests? Before you open new bug report, run emerge --sync. I think this issue was fixed three days ago but I can't say for sure without build.log.
Indeed, this issue was fixed a few days ago withe the "md2-optional" epatch from bug #331299 :). Case closed.
Arch teams, please, go ahead.
(In reply to comment #8) > Arch teams, please, go ahead. $ grep KEYWORDS *.ebuild ejabberd-2.0.5-r3.ebuild:KEYWORDS="amd64 ~arm ~ia64 ~ppc ~sparc x86" ejabberd-2.1.4.ebuild:KEYWORDS="amd64 ~arm ~ia64 ~ppc ~sparc x86" ejabberd-2.1.5.ebuild:KEYWORDS="amd64 ~arm ~ia64 ~ppc ~sparc x86" hrm?
Err, looks like I've pushed it to stable together with openssl-1.x fix. My bad :( : $ cvs diff -r 1.1 ejabberd-2.1.5.ebuild --- ejabberd-2.1.5.ebuild 5 Aug 2010 10:41:17 -0000 1.1 +++ ejabberd-2.1.5.ebuild 9 Aug 2010 08:21:28 -0000 1.2 @@ -14,7 +14,7 @@ -KEYWORDS="~amd64 ~arm ~ia64 ~ppc ~sparc ~x86" +KEYWORDS="amd64 ~arm ~ia64 ~ppc ~sparc x86" @@ -42,6 +42,7 @@ + epatch "${FILESDIR}/${P}-md2-optional.patch" #331299 Since I know for sure it works on amd64 I think amd64 is done here. x86 could you do your checks anyway, please? I don't think reverting keywords helps situation here.
Fine.