Re: darwin works with small tweak

[Rick Robino]

> On Thu, Sep 18, 2003 at 12:59:04AM +0200, Martin Pala wrote:
> Rick Robino wrote:
> 
> >     - the configure 'Check for ssl' includes a test for libraries ending
> >     in .dylib (only checks for .so and .a now)
> >
> I have added the test to configure.ac - i hope it will work for Darwin 
> (i cannot test it)

Works.

> >     - [ssl only] system paths /usr/include and /usr/lib are not fed to
> >     gcc.
> >
> I can't test it - if you can deliver patch, we can add it to the 
> distribution.

--- configure.ac.orig   Wed Sep 17 17:14:32 2003
+++ configure.ac        Wed Sep 17 17:33:06 2003
@@ -433,8 +433,13 @@
 # Add SSL includes and libraries
 if test "$sslincldir" -a "$ssllibdir"
 then
-    CFLAGS="$CFLAGS -I$sslincldir"
-    LIBS="$LIBS -L$ssllibdir -lssl -lcrypto"
+    # Darwin aleady knows about ssldirs
+    if test "x$ARCH" = "xDARWIN"; then
+     LIBS="$LIBS -lssl -lcrypto"
+    else
+     CFLAGS="$CFLAGS -I$sslincldir"
+     LIBS="$LIBS -L$ssllibdir -lssl -lcrypto"
+    fi
     # Redhat 9 compilation fix:
     if test -f "/usr/kerberos/include/krb5.h"
     then

> >The malloc quirk remains with cli status checks, whether or not ssl
> >support is built, but I think it's safe to ignore because status
> >is ephemeral and doesn't affect the daemon. A warning in the platforms
> >file would suffice, imho. 
> >
> I'll look on it (if you will assist :) I saw backtrace, etc. which you 
> send in the past - we must find the exact place where it failed. In the 
> case that it will be possible to fix it in monit (i hope so), we can 
> include the patch to the nearest release after 4.0.

As I mentioned in that past letter I can make the error go away by
adjusting just one expression in socket.c; I don't grok the macros
and structs used in monit well enough to be certain why this works
ok but that the best indication that I have of *where* the nit may
be found.

You can email me offline with any instructions, I'll assist any way
you want.

> >SSL did actually work in my tests - which were on b4 and b5.  It
> >built but didn't seem to work before then.

The patch you added and my patch do result in a proper compile
directive that builds ssl support without warnings, but with the
latest cvs I get an error somewhere else so I can't actually test
the end result.  FWIW, here is my current error:

gcc -c -g -O2 -Wall -DREENTRANT -no-cpp-precomp -DNEED_SOCKLEN_T_DEFINED 
-DDARWIN  -I. -I./device -I./http -I./process -I./protocols http/engine.c -o 
http/engine.o
http/engine.c: In function `socket_producer':
http/engine.c:462: storage size of `fds' isn't known
http/engine.c:466: `POLLIN' undeclared (first use in this function)
http/engine.c:466: (Each undeclared identifier is reported only once
http/engine.c:466: for each function it appears in.)
http/engine.c:466: `POLLPRI' undeclared (first use in this function)
http/engine.c:470: warning: implicit declaration of function `poll'
http/engine.c:462: warning: unused variable `fds'
make: *** [http/engine.o] Error 1

-- 
Rick Robino