Martin Knoblauch wrote:
Hi Matt, Folks,

 you may have seen the commits. I have just moved "libmetrics" out of
"scrlib" into top-level. "srclib" is now external-only stuff.

 Build tested on my FC4 system. I will try HP-UX and Solaris later the
week.

that's a good idea.  makes more sense.

talking about library dependencies..

if we look in ./gmond/Makefile.am you will see...

gmond_LDADD   = $(top_builddir)/libmetrics/libmetrics.la \
                $(top_builddir)/lib/libganglia.la \
                $(top_builddir)/lib/libgetopthelper.a \
                $(top_builddir)/srclib/confuse/src/libconfuse.la \
                $(top_builddir)/srclib/apr/libapr-0.la
gmond_LDFLAGS = -static

there are the dependencies that gmond has. of course we need libmetrics and libganglia (we wrote them:)). the libgetopthelper library is there for systems that don't have getopt() and its friends. libconfuse is the configuration file parser. last but not least is libapr (for portable networking, hash tables, signals, dso (future), arrays, i/o, etc..).

future work:

* we could remove libgetopthelper altogether since apr has the same functions...
<http://apr.apache.org/docs/apr/group__apr__getopt.html>
shouldn't be too painful.. maybe i'll take a look at this soon.

the gmetad dependencies in ./gmetad/Makefile.am are...

gmetad_LDADD   = $(top_builddir)/lib/libganglia.la -lrrd -lm \
                 $(top_builddir)/srclib/expat/lib/libexpat.la
gmetad_LDFLAGS = -static

gmetad basically only needs librrd (rrdtool) and libexpat (xml parser).

i didn't mean to say before that we should only support static linking of these libraries. i know i used the phrase "critically important" (a bit overstated). i just want to make sure that the default behavior stays close to what we have now. ganglia has been around now for over 5 years and people have installation routines that i don't want to change out from under them.

btw, has anyone done a "ganglia cluster report" search on google? it returns 450,000 results. i'm only able to see the first 1000 but they are all ganglia pages. i emailed google to see if there is a way to see if there really are half a million ganglia pages out there. i doubt it but i'd love to know the real number.

-matt

Reply via email to