httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Ralf S. Engelschall" <>
Subject Re: NO RELEASE.......................... (was RE: STATUS!)
Date Sat, 20 Mar 1999 18:46:57 GMT

In article <> you wrote:
> According to Lars Eilebrecht:
>>      * mod_access/3821: check_dir_access(), reads a->order[31]
>>        => either an if-clause should be added or the M_INVALID changed
>>      * mod_so/3493: os/unix/os.c dlclose()s objects before module cleanups
>>        => this is a subtle thing related to pool cleanups, I think.
> Ok, IMO these are our showstoppers and MUST BE fixed for 1.3.5.
> I won't roll the tarball until these bugs are fixed!!
> Dean, please look at the check_dir_access() problem. AFAIK you have the
> most experience with the directory walk stuff.

I've now recognized from the logs that the M_INVALID define Roy introduced
when he added the DAV related stuff, AFAIK. So he seems to be the guy who
knows whether its cleaner to change the value of M_INVALID or introduce an
if-clause in mod_access.

> I just talked to Ralf on the phone and he told me that he'll try
> to look at the dlclose() problem, but someone else should at it, too.
> Martin, how about you?

As I've now notices, Ben H. already investigates some time on this, too. And
from his explanation in the BugDB it looks that this is a non-problem. At
least when it exists, it seems not really causes by some wrong cleanup
procecure. The LIFO principle is correct here and should work. Ben H., are you
convinced that we've no problem here. I now finally cannot find a problem
myself, too.

>>    * mod_rewrite/3874: RewriteLock doesn't work for virtual hosts 
>>      => When I find time, I can look at this. But I would appreciate
>>         when someone other already can dive into this. My opinion
>>         is already appended to the PR.
> Ralf may not has the time to fix this for 1.3.5. Can someone else
> fix this bug??

A real solution for this flock-related problem seems to need more code.  And
this is too risky for 1.3.5. So, I think we should live with the current state
for 1.3.5, unless I was able to trace down and solve the problem the PR
describes. It's harmless, because flock is only one implementation for
RewriteLock and RewriteLock is only needed for RewriteMap and even when
not available more than garbled map lookups cannot happen.

                                       Ralf S. Engelschall

View raw message