Showstopper for 1.3.33

[Jim Jagielski <ji...@jaguNET.com>]

There is currently one showstopper holding up release of 1.3.33.
It has 2 votes for and none against, and it's for backing out
a patch recently applied in mod_rewrite... Please look it
over, otherwise I'll assume a lazy consensus (I know, I know)
and apply it.

Re: Showstopper for 1.3.33

[Jan Kratochvil <rc...@jankratochvil.net>]

On Wed, 27 Oct 2004 11:38:14 +0200, Joe Orton wrote:
> On Wed, Oct 27, 2004 at 11:11:44AM +0200, Jan Kratochvil wrote:
...
> > http://issues.apache.org/bugzilla/show_bug.cgi?id=14518
...
> > The patch during testing appears to me as inappropriate for 1.3.x; backout
> > is IMO appropriate.
> > The patch was right for 2.0.x (hopefully the backport to 1.3.x is not mine...).
> 
> Why do you think it's appropriate for 2.0.x?  The patch also breaks [P]
> with a query string in 2.0.x; at least in configurations which work
> working fine previously by my testing.

In such case going to more thoroughly test the 2.0.x behavior. It fixed
my mod_rewrite+mod_proxy case of use shown in my Bug comment 2004-10-25 00:52.

I think the real bug is in proxy_http_canon() and mod_rewrite() is just
a workaround of it. Unfortunately I am not sure what are the right inputs and
outputs of proxy_http_canon().


Regards,
Lace

Re: Showstopper for 1.3.33

[Joe Orton <jo...@redhat.com>]

On Wed, Oct 27, 2004 at 11:11:44AM +0200, Jan Kratochvil wrote:
> Hi,
> 
> On Tue, 26 Oct 2004 14:13:08 +0200, Jim Jagielski wrote:
> > There is currently one showstopper holding up release of 1.3.33.
> 
> http://issues.apache.org/bugzilla/show_bug.cgi?id=14518
> 
> > It has 2 votes for and none against, and it's for backing out
> > a patch recently applied in mod_rewrite...
> 
> The patch during testing appears to me as inappropriate for 1.3.x; backout
> is IMO appropriate.
> The patch was right for 2.0.x (hopefully the backport to 1.3.x is not mine...).

Why do you think it's appropriate for 2.0.x?  The patch also breaks [P]
with a query string in 2.0.x; at least in configurations which work
working fine previously by my testing.

joe

Re: Showstopper for 1.3.33

[Jan Kratochvil <rc...@jankratochvil.net>]

Hi,

On Tue, 26 Oct 2004 14:13:08 +0200, Jim Jagielski wrote:
> There is currently one showstopper holding up release of 1.3.33.

http://issues.apache.org/bugzilla/show_bug.cgi?id=14518

> It has 2 votes for and none against, and it's for backing out
> a patch recently applied in mod_rewrite...

The patch during testing appears to me as inappropriate for 1.3.x; backout
is IMO appropriate.
The patch was right for 2.0.x (hopefully the backport to 1.3.x is not mine...).


Still it is just a test by heuristics as I cannot check the code itself.
I am clueless what is the expected value of URI-parsing fields of
"struct request_rec":
unparsed_uri, uri, filename, canonical_filename, path_info, args, parsed_uri.

All these items I consider undocumented - both in 1.3 and in 2.1 trees
(comments in 2.1 are still very ambiguous - URI examples would be perfect).

Unfortunately this behavior is pretty hot for modules compatibility.
What about cleaning it out before 2.2?
Signature of proxy_http_canon() looks broken to me:
 * Its "url" argument type is undocumented as in the cases above.
 * Argument "url" relation to "r" is also undocumented.
 * The function modified its "url" argument content - IMO a bit unusual.
 * The 'r->uri == r->unparsed_uri' code path condition looks unclear to me,
   there should be at least macro for such condition defined around
   "struct request_rec".

I would not say anything regarding undocumented URL code/structures if these
URL parsing issues would be handled fine by the code aware httpd coders;
unfortunately they look to have the same problems as httpd newbie like me.


Regards,
Lace