README v1.5 - darwin build note

[Rick Robino]

Greetings hauk, long time no talk to.  This is a developer type 
message, but the general list was all I could find.

I happened to be peeking about the CVS repository for monit to see what 
is cooking for v4 and read that the Mac OS X port was giving you 
problems on the sf.net compile farm.  Since your code is so beautiful I 
just knew it would compile with only a little autoconf push ;-)   
Actually, I tried to reproduce the pthread sigmask problem and didn't 
see it in the standard error output from configure/make, but I did get 
some problems about multiple and undeclared  definitions of FALSE and 
TRUE, etc. - most apple folks by now are familiar with these sorts of 
silly errors.  Addressing two items in configure(.ac) yields a working 
binary for me, with just a few innocuous warnings:

1.  Apple's linker is kind of strange, and the behavior has changed 
between major versions, so some things that didn't generate complaints 
now do and vice versa - the complaints do not necessarily indicate a 
real problem that the compiler is saving you from.  For example:  
multiple definitions.  Anyway, the magic incantation with the latest 
version of the tools Apple ships is:  -multiply_defined 
<error|warn|suppress>.  Default is error, but suppressing them is 
usually safe if you think it should be.  FWIW, the other common bogus 
compiler error tends to be cured with: CFLAGS="-no-cpp-precomp", just 
in case you or someone else run into any other common mac problems.

2.  Since gcc is Apple's native compiler, AND openssl is actually part 
of the system (as in residing in /usr instead of /usr/local) the ssl 
tests pick up system paths that gcc has already fixed the includes for, 
causing gcc to choke on the new -I or -L that is already part of its 
standard lookup path.  OpenBSD has a similar ssl situation, I would 
imagine that the current configure runs into this same problem on that 
platform as well.

If #1 is not handled the kit still compiles but with the 
multiply_defined warning, but #2 is a show stopper for gcc if ssl 
support is built.  Once resolved, the only warnings remaining are 
during the final stages where gcc complains:

gcc -c -g -O2 -Wall -DREENTRANT -DDARWIN   -I. -I./protocols -I./http 
-I./process  lex.yy.c -o lex.yy.o
lex.yy.c:2494: warning: `yyunput' defined but not used
gcc -c -g -O2 -Wall -DREENTRANT -DDARWIN   -I. -I./protocols -I./http 
-I./process  process/sysdep_DARWIN.c -o process/sysdep_DARWIN.o
process/sysdep_DARWIN.c: In function `calcru':
process/sysdep_DARWIN.c:103: warning: unused variable `tv'
process/sysdep_DARWIN.c: In function `get_process_info_sysdep':
process/sysdep_DARWIN.c:215: warning: unused variable `pstats'
process/sysdep_DARWIN.c:218: warning: unused variable `rup'
/usr/include/stdio.h: At top level:
process/sysdep_DARWIN.c:96: warning: `calcru' defined but not used

Upon running the finished product the runtime linker (?) does give a 
warning:

$ ./monit -c ~/.monitrc -d 60
*** malloc[3867]: Deallocation of a pointer not malloced: 0x2359a4; 
This could be a double free(), or free() called with the middle of an 
allocated block; Try setting environment variable MallocHelp to see 
tools to help debug
Starting monit daemon
Starting httpd at [localhost:2812]

I haven't looked any farther into this warning, other than testing out 
the running daemon - seems to work well with my fairly complex config.  
Service restarts work, monit takes signals and does not die, for the 
short time I ran the daemon I did not see it grow stack size in top(1). 
 So... unless there really is a free() bug or the duplicate symbol 
truly does pose a problem, I'd say that monit does compile and work 
fine on OS X.  Very tiny configure.ac diff attached.

My environment particulars should be common to most present mac users, 
thanks to auto update:
Darwin spiff.local. 6.6 Darwin Kernel Version 6.6: Thu May  1 21:48:54 
PDT 2003; root:xnu/xnu-344.34.obj~1/RELEASE_PPC  Power Macintosh powerpc
(a.k.a Mac OS X 10.2.6)

Apple Developer Tools, December 2002:
gcc (GCC) 3.1 20020420 (prerelease)
autoconf 2.52  (yes, I fudged this one down in configure.ac).


Cheers,

--
Richard T. Robino
Internet Systems Architect
Wave Division Consulting
http://www.wavedivision.com/~rrobino/