getBaseUrl.py doesn't work as expected behind Apache virtual hosts

I need more rules to understand the whole bug. Here my generic rule: RewriteRule? ^{/(.*) \ http://localhost:5080/VirtualHostBase/http/%{HTTP_HOST}:80/cps/VirtualHostRoot/somewhere/$1 [L,P]}

It transforms domain.com/somewhere/ in localhost:5080/cps/.

Here is what I use:

ServerName? mysite.com RewriteEngine? on

RewriteCond? %{HTTP:Authorization} ^{(.*) RewriteRule?} /(.*) http://localhost:5080/VirtualHostBase/http/%{HTTP_HOST}:80/cps/VirtualHostRoot/$1 [L,P]

The only difference seems to be the presence of "somewhere" at the end of the rewrite rule.

From now on, we will include the Apache httpd rewriting rule to be used with CPS with each CPS release.

To examine also harder to find problems with virtual hosting, I set up the following hierarchy of folders/cps instances in zope:

root of zope

! +--cps1 ! ! +--folder--+

! +--cps2 ! +--index_html ! +--subfolder--+

! +--cps3

To check it, I tried to reach each 'reachable' cps instance defined by any of the rules below

The following 12 rewriting rules apply to the above pattern:

# case 1:

ServerName? zope1.pu.smp.net RewriteEngine? On RewriteRule? ^{/(.*) \}

http://127.0.0.1:8280/VirtualHostBase/http/zope1.pu.smp.net:80/VirtualHostRoot/$1

[L,P] ProxyVia? on

# case 2:

ServerName? cps1.pu.smp.net RewriteEngine? On RewriteRule? ^{/(.*) \}

http://127.0.0.1:8280/VirtualHostBase/http/cps1.pu.smp.net:80/cps1/VirtualHostRoot/$1

[L,P]

ProxyVia? on

# case 3:

ServerName? zope2.pu.smp.net RewriteEngine? On RewriteRule? ^/(.*)

http://127.0.0.1:8280/VirtualHostBase/http/zope2.pu.smp.net:80/folder/VirtualHostRoot/$1 [L,P]

ProxyVia? on

# case 4:

ServerName? cps2.pu.smp.net RewriteEngine? On RewriteRule? ^/(.*)

http://127.0.0.1:8280/VirtualHostBase/http/cps2.pu.smp.net:80/folder/cps2/VirtualHostRoot/$1 [L,P]

ProxyVia? on

# Inside-out hosting with single subdirectory... # case 5:

RewriteRule? ^{/var/www/sub1/(.*) http://127.0.0.1:8280/VirtualHostBase/http/pu.smp.net:80/VirtualHostRoot/_vh_sub1/$1 [L,P]}

# case 6:

RewriteRule? ^{/var/www/sub2/(.*) http://127.0.0.1:8280/VirtualHostBase/http/pu.smp.net:80/cps1/VirtualHostRoot/_vh_sub2/$1 [L,P]}

# case 7:

RewriteRule? ^{/var/www/sub3/(.*) http://127.0.0.1:8280/VirtualHostBase/http/pu.smp.net:80/folder/VirtualHostRoot/_vh_sub3/$1 [L,P]}

# case 8:

RewriteRule? ^{/var/www/sub4/(.*) http://127.0.0.1:8280/VirtualHostBase/http/pu.smp.net:80/folder/cps2/VirtualHostRoot/_vh_sub4/$1 [L,P]}

# Inside-out hosting with multiple subdirectories... # case 9:

RewriteRule? ^{/var/www/sub5/sub1/(.*) http://127.0.0.1:8280/VirtualHostBase/http/pu.smp.net:80/VirtualHostRoot/_vh_sub5/_vh_sub1/$1 [L,P]}

# case a: RewriteRule? ^{/var/www/sub5/sub2/(.*)}

http://127.0.0.1:8280/VirtualHostBase/http/pu.smp.net:80/cps1/VirtualHostRoot/_vh_sub5/_vh_sub2/$1 [L,P]

# case b: RewriteRule? ^{/var/www/sub5/sub3/(.*)}

http://127.0.0.1:8280/VirtualHostBase/http/pu.smp.net:80/folder/VirtualHostRoot/_vh_sub5/_vh_sub3/$1 [L,P]

# case c: RewriteRule? ^{/var/www/sub5/sub4/(.*)}

http://127.0.0.1:8280/VirtualHostBase/http/pu.smp.net:80/folder/cps2/VirtualHostRoot/_vh_sub5/_vh_sub4/$1 [L,P]

Uli Fouquet

_vh_sub5/_vh_sub2/$1 does not make much sense

why not write:

_vh_sub5/sub2/$1 ?

the doc says "VirtualHostRoot?, followed by one or *more* path elements that start with '_vh_'":

If you insert VirtualHostRoot?, followed by one or more path elements that start with '_vh_', then these elements will be ignored during traversal and then added (without the '_vh_') to the start of generated URLs. For instance, a request for "/a/VirtualHostRoot/_vh_z/" will traverse "a" and then generate URLs that start with /z.

In our example, you would have the main server send requests for http://www.mycause.org/dynamic_stuff/anything to Zope, rewritten as /VirtualHostRoot/_vh_dynamic_stuff/anything.

_

it should be enough to have:

if path_info.startswith('/VirtualHostBase/'):

# apache detection

# Inside-out hosting (VHM _vh_) if path_info.find('_vh_') > 0:

base = path_info.split('_vh_')[1] return '/' + base

else:

return '/'

and remember to set the trailing '/' in the rewrite rule.

There is a difference between, let's say

_vh_sub5/cps/$1

and

_vh_sub5/_vh_cps/$1

The first expression assumes 'cps' is an existant zope object while the second one assumes it to be a virtual apache directory.

First one should return the object $1 located in a folderish zope object called cps, while the second one will return the zope object $1.

This is obvious when you watch the "VIRTUAL_URL_PARTS" variable after using both of the rules:

First one will result in something like:

('http://localhost', 'sub5', 'cps/$1')

while second one will give:

('http://localhost', 'sub5/cps', '$1' )

The docs of course can be misunderstood:

"followed by one or more path elements that start

with '_vh_'"

means

"followed by one or more path elements that start

with '_vh_' *each*"

You must have a chance to mark a difference between (existant) subdirectories in zope and (virtual) subdirectories in apache.

Therefore looking for the first occurence of '_vh_' might not be enough. There may be more of them and that with good reason.

OK on the principle, but what difference is there between:

('http://localhost', 'sub5', 'cps/$1')

and:

('http://localhost', 'sub5/cps', '$1' )

when you apply '/'.join() to them which I am sure VHM does a few lines later.

what are the use cases where the distinction is important because, independently of the internal computations used, the idea is to translate a URL into another URL?

I see your point.

I think, in first case the BaseUrl? should be '/cps/' while in second case it should be '/'.

This makes a difference for scripts building links (like getBreadCrumbs()).

This bug is not blocking anymore since Encolpe has done a rollback on both HEAD and CPS-3_2-branch branches.

Fixed in [25046]. Some issues still need to be explored, see #847