Notes from building the 2.0.15 tarball - in case anyone is interested.

The link on the top of http://httpd.apache.org/dist/ 
and http://httpd.apache.org/dist/httpd should be:

http://httpd.apache.org/dist/httpd/Announcement2.html

(not http://httpd.apache.org/dist/Announcement2.html)
---
Typo on the bottom of http://httpd.apache.org/dist: 
2.0i.15 should be 2.0.15

The following build was on a Mandrake 7.1 (Linux 2.2) box.  This isn't my 
standard development machine, so this is as close to a virgin box as we'll 
find around here.  And, it is also a different platform than what I
traditionally build with.

"./configure --prefix=/where/i/wanted/it/to/go; make; make install" 

So, there should be nothing special about this build or setup - straight
out of the tarball.

The APR_CHECK_ICONV_INBUF test in apr printed its warning message to stderr 
(Mandrake 7.1 has const char **-based iconv).  I wonder if it would be 
possible to redirect the output of this warning from stderr to 5 (which seems 
to go into the config.log file).  I seem to remember a thread about compiler 
warnings a while ago, but I don't remember the conclusion.  I don't think
warning should go away (obviously it serves a purpose here) - it might just 
be redirected to somewhere where it can be recorded by the build system.
---
Apache is set to have the Port default to 80, but Mandrake already has an 
Apache instance running on port 80 by default.  When I tried to start 2.0.15
without realizing there was already another server running, apachectl 
said that the 2.0.15 httpd started, but it didn't "really" start because of 
the httpd already running on port 80 (obviously the bind failed).  Of course 
changing the Port directive to something else worked fine...

It seems like the more intuitive thing would be to delay exit of apachectl
until we know that we successfully started httpd.  This might mean delaying 
the double-fork of httpd or something similar, so this might get tricky.  Or, 
we could add something to apachectl (i.e. sleep a few seconds and see if our 
pidfile still exists).  I'm not sure - I don't think the naive (or rushed) 
admin would expect to check for this case if apachectl said it started it 
without realizing it didn't start because there was another server running
that it didn't control.  Yes, this might fall under user error, but still 
something that might be worth considering to make apachectl or httpd report 
the right thing.

No show stoppers just some minor nitpicks (really more feature-enhancements).
I'm not going to perform any load tests on this setup.  But, it looks good.
I know that the current CVS tree on my Linux 2.4/Intel box and our 
Solaris 7/Sparc box are reasonably stable...

So, from my non-committer (hence almost worthless) standpoint, +1 for getting
this build out to a wider audience.

HTH,
Justin Erenkrantz
[EMAIL PROTECTED]

Reply via email to