[Moin-user] Odd bug related to If-Modified-Since using moin 1.5.4 with CGI on apache

[Thomas Waldmann]

> There doesn't seem to be a stack trace generated.
Sorry, I misread that in your first msg.

> The affected moin instances aren't returning Cache-control:no-cache,
> Pragma:no-cache and Expires:-1.
As I said, I found an error concerning setting of header values 
recently. That error might make some headers missing for some server 
types (esp. Vary and Cache-Control - because those used to do multiple 
setHeader calls, which didn't work for every server type. The 
non-working types just emitted the value from the last call).

But: pure standard CGI didn't have that header problem.
> Also, it seems newer moinmoin instances are now returning Last-Modified,
Yes.

But, as I found out recently, they do wrong for some dynamic pages like 
RecentChanges (they use the mtime of the page src file, which never 
changes, but RecentChanges output changes all the time). This has still 
to get fixed.

It is no problem if the user is logged in, but anon users might see a 
stale RecentChanges.
> which they previously didn't, this would mask any bug regarding
> If-Modified-Since: that would never even be sent.
So here you tell it doesn't happen with 1.5.4?
> In summary:
> - it chokes on If-Modified-Since
> - a python stack trace doesn't seem to be generated (but is generated on
> other faults)
> - some caching related headers seem to missing from the response
Strange. CGI should not be affected. You are sure you don't use 
mod_python or fastcgi for moin?

> - I can't find another moinmoin wiki on the web suffering from the same
> problem, so it's probably(?) local - then again I can't find that many
> moinmoin 1.5.4's ;-)
Here you tell it happens with 1.5.4???

moinmoin.wikiwikiweb.de currently runs on 1.5.4+ (code might be newer, 
but not older than 1.5.4) under Twisted.
> so:
> -it might be a config error, however to my knowledge with the 
> exception of
> the .htaccess hacks I use to fix up the situation nothing 
> caching-wise, or
> header-wise for that matter is different from the default.
> -it might be a moinmoin bug, but then almost certainly CGI only; as there
> are quite a few non-CGI wikis online and none of them seem to be affected
> (but almost all of those I tested do obviously custom things with 
> headers).
> It would make sense for header errors to be different on different 
> platforms
> as it doesn't surprise me that a CGI served wiki deals with those
> differently than a twisted or standalone or mod_python or whatnot server
> version would.
>
> If none of this rings any bells I'll just wait until 1.6 and investigate
> more thoroughly then.
I guess I backport the header fixes I did in 1.6 soon - I could give you 
a patch for 1.5.4 then to try.