I think I have it figured out now; I am posting what I have found to the
list in the hopes of pointing out some potential problems.
I'm running redhat 9.0 and wanted to install Subversion for evaluation
purposes. I'll spare you the details, here are the important points.
1. RH 9.0 binary RPMS didn't work:
http://summersoft.fay.ar.us/pub/subversion/latest/redhat-9.0/
(linked from: http://subversion.tigris.org/project_packages.html)
"svnadmin create.. " gave some sort of Berkeley db error.
("Operation not supported" ?)
installing newer berkeley db didn't help.
2. I read that installing from sources would fix this problem.
(got the src.rpm from the same summersoft site)
HERE IS THE FIRST BIGGIE. This src rpm doesn't include
all the apr sources (or configure script) as I'm told it should.
3. This led me down the wrong path of running ./autogen.sh
( I KNOW...I'm not supposed to run that, but the source rpm
didn't have a configure script...) and trying to install new
versions of apr (which install themselves in nonstandard
locations, evidently...causing ./configure problems)
(I think apr installed itself into /usr/local/apr by default.
I believe the ./configure script doesn't look there -- instead
it finds the one that came with RH9 in /usr (*I THINK*) )
4. Finally when it all seems in place, the "Apache" problem comes up.
With both the "munged install process" *and* the source tarball
process, configure gives this error:
checking httpd version... configure: error: apache too old: mmn must be
at least 20020903
I repeated this test on a different RH9 machine, and did not get this
error.
My suspicion is that "Old Apache" is worse than "No Apache". I can't say
I really follow what the configure script is doing, but I suspect this
might be the problem:
configure:8849
if test -r $APXS_INCLUDE/mod_dav.h; then
...
if (eval "$ac_cpp conftest.$ac_ext") 2>&5 |
$EGREP "VERSION_OKAY" >/dev/null 2>&1; then
...
else { ... (exit 1);
if the first "if" passes, (it found some sort of apache headers), then
the second test is issued (check for recent version). If the version is
too old, the script complains and exits. If no headers were found in the
first place, then the script continues on it's way.
So in my case, it found an old version of apache and puked.
I say either this test shouldn't fail in this way (ie old apache gets
treated the same as no apache) or the error message be more clear.
I ended up having to use --with-apache=no --with-apxs=no
(at least I think that is what fixed it)
-Steve
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Thu Sep 16 06:38:54 2004