Static gitweb content when using pathinfo

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hi,

I'm running gitweb in a server.example.com/gitweb/ scenario.  If I don't enable
pathinfo everything works fine, but when I enable pathinfo the static content
stops showing up in the browser.

I'm not an HTML or perl expert, but I think that this may be due to a missing
slash in the construction of the base tag:  Using the upstream script I am getting

<base href="http://server.example.com/gitweb" />

and no static content.  Once I tune the script to add a final slash to the url

<base href="http://server.example.com/gitweb/" />

the static content shows up again.

Maybe I'm doing something wrong (am I?), and possibly this can be solved from
the web server side, but maybe it also makes some sense to include in the script
some check that the url used in the base tag ends with a slash.  In that case,
given my lack of perl skills, I would be really grateful if somebody implemented it.

Thank you for your attention.

Greetings,

	E.

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.11 (GNU/Linux)

iEYEARECAAYFAk99pBMACgkQY+7weQMem3y28wCeOotCIgFF8sT4LKgq599IhB/E
0qcAn2dri0Z61PjV90bxfBG1QMFD3c75
=6PBz
-----END PGP SIGNATURE-----

On Thu, 5 Apr 2012, José María Escartín Esteban wrote:

Hi,

I'm running gitweb in a server.example.com/gitweb/ scenario.  If I don't enable
pathinfo everything works fine, but when I enable pathinfo the static content
stops showing up in the browser.

I'm not an HTML or perl expert, but I think that this may be due to a missing
slash in the construction of the base tag:  Using the upstream script I am getting

<base href="http://server.example.com/gitweb" />

and no static content.  Once I tune the script to add a final slash to the url

<base href="http://server.example.com/gitweb/" />

the static content shows up again.

Maybe I'm doing something wrong (am I?), and possibly this can be solved from
the web server side, but maybe it also makes some sense to include in the script
some check that the url used in the base tag ends with a slash.  In that case,
given my lack of perl skills, I would be really grateful if somebody implemented it.

How do you deploy gitweb, what is your web server configuration, and
what is the URL of main gitweb page (using placeholders like 'example.com'
or 'foo')?

When I am running gitweb, it uses http://localhost/cgi-bin/gitweb/gitweb.cgi
URL, and BASE element for path_info is

  <base href="http://localhost/cgi-bin/gitweb/gitweb.cgi" />

Note that last component of 'href' attribute of base element is the
basename of script, which according to HTML 4.01[1] and RFC 3986[2]
is removed[3]

   o  return a string consisting of the reference's path component
      appended to all but the last segment of the base URI's path (i.e.,
      excluding any characters after the right-most "/" in the base URI
      path, or excluding the entire base URI path if it does not contain
      any "/" characters).


There might be problem if you configured your web server to serve gitweb
using it as a handler for subdirectory, so the script name does not need
to appear in URL, e.g.

  http://localhost/cgi-bin/gitweb

which would require the following base element

  <base href="http://localhost/cgi-bin/gitweb/" />

Gitweb would have to distinguish those two cases somehow, and either add
or not add trailing slash '/'.

[1]: http://www.w3.org/TR/html401/struct/links.html#h-12.4
[2]: http://tools.ietf.org/html/rfc3986#section-5.2
[3]: http://tools.ietf.org/html/rfc3986#section-5.2.3

Once I tune the script to add a final slash to the url

<base href="http://server.example.com/gitweb/" />

the static content shows up again.

By "tune the script" do you mean editing gitweb.cgi, or adding setting
$base_url to gitweb configuration file, see gitweb.conf(5):

  $base_url
     Base  URL  for  relative  URLs  in  pages generated by gitweb, (e.g.
     $logo, $favicon, @stylesheets if they are relative URLs), needed and
     used  <base href="$base_url"> only for URLs with nonempty PATH_INFO.
     Usually gitweb sets its value correctly, and there is no need to set
     this  variable,  e.g.  to $my_uri or "/". See $per_request_config if
     you need to override it anyway.

You can e.g. put

  $base_url .= '/' unless ($base_url =~ m!/$!);

in the gitweb configuration file to ensure that it ends with '/', whatever
it is.

HTH
-- 
Jakub Narebski
Poland

On Thu, 5 Apr 2012, Jakub Narebski wrote:

On Thu, 5 Apr 2012, José María Escartín Esteban wrote:

quoted

Hi,

I'm running gitweb in a server.example.com/gitweb/ scenario.  If I don't enable
pathinfo everything works fine, but when I enable pathinfo the static content
stops showing up in the browser.

I'm not an HTML or perl expert, but I think that this may be due to a missing
slash in the construction of the base tag:  Using the upstream script I am getting

<base href="http://server.example.com/gitweb" />

and no static content.  Once I tune the script to add a final slash to the url

<base href="http://server.example.com/gitweb/" />

the static content shows up again.

How do you deploy gitweb, what is your web server configuration, and
what is the URL of main gitweb page (using placeholders like 'example.com'
or 'foo')?

Ping?

[...]

There might be problem if you configured your web server to serve gitweb
using it as a handler for subdirectory, so the script name does not need
to appear in URL, e.g.

  http://localhost/cgi-bin/gitweb

which would require the following base element

  <base href="http://localhost/cgi-bin/gitweb/" />

[...]

quoted

Once I tune the script to add a final slash to the url

<base href="http://server.example.com/gitweb/" />

the static content shows up again.

By "tune the script" do you mean editing gitweb.cgi, or adding setting
$base_url to gitweb configuration file, see gitweb.conf(5):

  $base_url
     Base  URL  for  relative  URLs  in  pages generated by gitweb, (e.g.
     $logo, $favicon, @stylesheets if they are relative URLs), needed and
     used  <base href="$base_url"> only for URLs with nonempty PATH_INFO.
     Usually gitweb sets its value correctly, and there is no need to set
     this  variable,  e.g.  to $my_uri or "/". See $per_request_config if
     you need to override it anyway.

You can e.g. put

  $base_url .= '/' unless ($base_url =~ m!/$!);

in the gitweb configuration file to ensure that it ends with '/', whatever
it is.

By the way, would the following proposed addition to gitweb.conf(5)
manpage would help your situation?

-------- >8 ---------- >8 ---------
Subject: [PATCH/RFC] gitweb.conf(5): When to set $base_url

Add a paragraph to description of $base_url variable in gitweb.conf(5)
manpage explaining when and why one might need to set it.

Based-on-report-by: José María Escartín Esteban [off-list ref]
Signed-off-by: Jakub Narebski <redacted>
---
 Documentation/gitweb.conf.txt |   14 ++++++++++++++
 1 files changed, 14 insertions(+), 0 deletions(-)

diff --git a/Documentation/gitweb.conf.txt b/Documentation/gitweb.conf.txt
index 7aba497..a2a6ddf 100644
--- a/Documentation/gitweb.conf.txt
+++ b/Documentation/gitweb.conf.txt

@@ -559,6 +559,20 @@ $base_url::
 	PATH_INFO.  Usually gitweb sets its value correctly,
 	and there is no need to set this variable, e.g. to $my_uri or "/".
 	See `$per_request_config` if you need to override it anyway.
++
+You would need to set this variable when using gitweb as a directory
+handler and using path_info-based URLs.  For example if your web
+server is set up in such way that full path to browse repositories is
+`http://git.example.com/gitweb` and static files are served from
+'/gitweb/static' directory with default values (e.g. `$logo` is
+`static/git-logo.png`), then you would need to set `$base_url` to
+(for example):
++
+----------------------------------------------------------------------
+our $base_url = "http://git.example.com/gitweb/";
+----------------------------------------------------------------------
++
+The trailing slash is required!
 
 
 CONFIGURING GITWEB FEATURES

-- 
1.7.9

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hi,

Thank you for your detailed answers, and sorry for the delay in my reply.

On 14/04/12 19:19, Jakub Narebski wrote:

On Thu, 5 Apr 2012, Jakub Narebski wrote:

quoted

On Thu, 5 Apr 2012, José María Escartín Esteban wrote:

quoted

Hi,

I'm running gitweb in a server.example.com/gitweb/ scenario.  If I don't enable
pathinfo everything works fine, but when I enable pathinfo the static content
stops showing up in the browser.

I'm not an HTML or perl expert, but I think that this may be due to a missing
slash in the construction of the base tag:  Using the upstream script I am getting

<base href="http://server.example.com/gitweb" />

and no static content.  Once I tune the script to add a final slash to the url

<base href="http://server.example.com/gitweb/" />

the static content shows up again.

How do you deploy gitweb, what is your web server configuration, and
what is the URL of main gitweb page (using placeholders like 'example.com'
or 'foo')?

Ping?

[...]

I'll describe the setup at the end of my email, but you already provided the
solution!

quoted

There might be problem if you configured your web server to serve gitweb
using it as a handler for subdirectory, so the script name does not need
to appear in URL, e.g.

  http://localhost/cgi-bin/gitweb

which would require the following base element

  <base href="http://localhost/cgi-bin/gitweb/" />

[...]

Yes, that was precisely the problem.

quoted

quoted

Once I tune the script to add a final slash to the url

<base href="http://server.example.com/gitweb/" />

the static content shows up again.

By "tune the script" do you mean editing gitweb.cgi, or adding setting
$base_url to gitweb configuration file, see gitweb.conf(5):

  $base_url
     Base  URL  for  relative  URLs  in  pages generated by gitweb, (e.g.
     $logo, $favicon, @stylesheets if they are relative URLs), needed and
     used  <base href="$base_url"> only for URLs with nonempty PATH_INFO.
     Usually gitweb sets its value correctly, and there is no need to set
     this  variable,  e.g.  to $my_uri or "/". See $per_request_config if
     you need to override it anyway.

You can e.g. put

  $base_url .= '/' unless ($base_url =~ m!/$!);

in the gitweb configuration file to ensure that it ends with '/', whatever
it is.

I was editing manually the gitweb.cgi script to add (always!) the slash.  Your
solution is of course better, and completely solves my problem.

This addition to the man page would of course have been useful, since it took me
a while to detect that the problem was a missing slash in the base url.
However, it would not have solved completely my problem, and for me the best has
been to add the line of code you proposed in the previous email.


I am running a Debian testing (wheezy) server, so the distro gitweb files are:

	/etc/apache2/conf.d/gitweb
	/etc/gitweb.conf
	/usr/share/gitweb/gitweb.cgi
	/usr/share/gitweb/index.cgi -> /usr/share/gitweb/gitweb.cgi
	/usr/share/gitweb/static/*

I am using two sets of git repositories, and each of the sets works through a
particular gitolite user:

	user     home
	----------------------
	git      /srv/project
	mygit    /srv/mygit

I wanted that different groups of people were able to browse the 'project' repos
(in fact it is a mirror from the main project server) from
	http://server.example.com/project/ ,
and the 'mygit' repos from
	http://server.example.com/mygit/ .
And I wanted to keep urls as short and easy as possible, and to avoid
proliferation of configs as much as possible.

The setup I came up with was to link

	/srv/www/project -> /usr/share/gitweb
	/srv/www/mygit -> /usr/share/gitweb

and to use the following configuration files:

######### /etc/apache2/conf.d/gitweb ##################
<Directory /srv/www/project>
        SetEnv  GITWEB_PROJECTROOT      /srv/git/
        SetEnv  GITWEB_CONFIG           /etc/gitweb.conf
        <IfModule mod_authn_file.c>
                AuthType Basic
                AuthName "Project gitweb"
                AuthUserFile /srv/htpwd/git-htpasswd
        </IfModule>
        Require valid-user
        Order allow,deny
        Allow from 123.123.123.123
        Satisfy Any
        Options FollowSymLinks +ExecCGI
        AddHandler cgi-script .cgi
        RewriteEngine On
        RewriteBase /project/
        RewriteCond %{REQUEST_FILENAME} !-f
        RewriteCond %{REQUEST_FILENAME} !-d
        RewriteRule ^.* /project/index.cgi/$0 [L,PT]
</Directory>

<Directory /srv/www/mygit>
        SetEnv  GITWEB_PROJECTROOT      /srv/mygit/
        SetEnv  GITWEB_CONFIG           /etc/gitweb.conf
        <IfModule mod_authn_file.c>
                AuthType Basic
                AuthName "My gitweb"
                AuthUserFile /srv/htpwd/mygit-htpasswd
        </IfModule>
        Require valid-user
        Order allow,deny
        Allow from 213.213.213.213
        Satisfy Any
        Options FollowSymLinks +ExecCGI
        AddHandler cgi-script .cgi
        RewriteEngine On
        RewriteBase /mygit/
        RewriteCond %{REQUEST_FILENAME} !-f
        RewriteCond %{REQUEST_FILENAME} !-d
        RewriteRule ^.* /mygit/index.cgi/$0 [L,PT]
</Directory>
#######################################################

######### /etc/gitweb.conf ############################
$projectroot = $ENV{'GITWEB_PROJECTROOT'}."repositories";
$git_temp = "/tmp";
$projects_list = $ENV{'GITWEB_PROJECTROOT'}."projects.list";
@diff_opts = ();
$feature{'pathinfo'}{'default'} = [1];
$feature{'highlight'}{'default'} = [1];
$projects_list_description_width = 50;
#######################################################

Probably there are better ways to implement this, but at least this seems to
work, once I have added

	$base_url .= '/' unless ($base_url =~ m!/$!);

to /etc/gitweb.conf .

Greetings, and thanks again,

	E.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.11 (GNU/Linux)

iEYEARECAAYFAk+MOEsACgkQY+7weQMem3w4NwCfYHyX+vtap/HvV7/8CTYu42W2
AbEAoLi2868tD6LGoFdisiplWf9vSdRx
=RAme
-----END PGP SIGNATURE-----

On Mon, 16 Apr 2012, José María Escartín Esteban wrote:

On 14/04/12 19:19, Jakub Narebski wrote:

quoted

On Thu, 5 Apr 2012, Jakub Narebski wrote:

quoted

On Thu, 5 Apr 2012, José María Escartín Esteban wrote:

quoted

quoted

There might be problem if you configured your web server to serve gitweb
using it as a handler for subdirectory, so the script name does not need
to appear in URL, e.g.

  http://localhost/cgi-bin/gitweb

which would require the following base element

  <base href="http://localhost/cgi-bin/gitweb/" />

[...]

Yes, that was precisely the problem.

Note that you won't have problem with `http://localhost` as base url
(without any directory), or in more realistic case using virtual host
rather than virtual directory.  The problem is that without '/' the
path part of $base_url is stripped of last element.

Just for completeness.

[...]

This addition to the man page would of course have been useful, since it took me
a while to detect that the problem was a missing slash in the base url.
However, it would not have solved completely my problem, and for me the best has
been to add the line of code you proposed in the previous email.

So how about the proposed updated addition at the end of this email?

I am running a Debian testing (wheezy) server, so the distro gitweb files are:

	/etc/apache2/conf.d/gitweb
	/etc/gitweb.conf
	/usr/share/gitweb/gitweb.cgi
	/usr/share/gitweb/index.cgi -> /usr/share/gitweb/gitweb.cgi
	/usr/share/gitweb/static/*

I am using two sets of git repositories, and each of the sets works through a
particular gitolite user:

	user     home
	----------------------
	git      /srv/project
	mygit    /srv/mygit

I wanted that different groups of people were able to browse the 'project' repos
(in fact it is a mirror from the main project server) from
	http://server.example.com/project/ ,
and the 'mygit' repos from
	http://server.example.com/mygit/ .
And I wanted to keep urls as short and easy as possible, and to avoid
proliferation of configs as much as possible.

I think that is this "two sets" that created configuration with a problem.
 

The setup I came up with was to link

	/srv/www/project -> /usr/share/gitweb
	/srv/www/mygit -> /usr/share/gitweb

and to use the following configuration files:

######### /etc/apache2/conf.d/gitweb ##################
<Directory /srv/www/project>
        SetEnv  GITWEB_PROJECTROOT      /srv/git/
        SetEnv  GITWEB_CONFIG           /etc/gitweb.conf

[...]

#######################################################

######### /etc/gitweb.conf ############################
$projectroot = $ENV{'GITWEB_PROJECTROOT'}."repositories";
$git_temp = "/tmp";
$projects_list = $ENV{'GITWEB_PROJECTROOT'}."projects.list";
@diff_opts = ();
$feature{'pathinfo'}{'default'} = [1];
$feature{'highlight'}{'default'} = [1];
$projects_list_description_width = 50;
#######################################################

A question: why not have

          SetEnv  GITWEB_PROJECTROOT      /srv/git

and use

  $projectroot = "$ENV{'GITWEB_PROJECTROOT'}/repositories";
  $projects_list = "$ENV{'GITWEB_PROJECTROOT'}/projects.list";
 

BTW. what is this $git_temp?  Is it modified gitweb, because core one
doesn't need to use temporary directory for anything nowadays?

A tip: you can use GITWEB_CONFIG_COMMON for the common part of
configuration and separate GITWEB_CONFIG for per-instance configuration.
Though I am not sure if it would help in your case.  It requires
modern gitweb. though.

Probably there are better ways to implement this, but at least this seems to
work, once I have added

	$base_url .= '/' unless ($base_url =~ m!/$!);

to /etc/gitweb.conf .

-- >8 ---------- >8 --
Subject: [PATCH] gitweb.conf(5): When to set $base_url

Add a paragraph to description of $base_url variable in gitweb.conf(5)
manpage explaining when and why one might need to set it, and how.

Based-on-report-by: José María Escartín Esteban [off-list ref]
Signed-off-by: Jakub Narębski <redacted>
---
 Documentation/gitweb.conf.txt |   24 ++++++++++++++++++++++++
 Documentation/gitweb.txt      |    4 ++++
 2 files changed, 28 insertions(+), 0 deletions(-)

diff --git a/Documentation/gitweb.conf.txt b/Documentation/gitweb.conf.txt
index 7aba497..4716a0f 100644
--- a/Documentation/gitweb.conf.txt
+++ b/Documentation/gitweb.conf.txt

@@ -559,6 +559,30 @@ $base_url::
 	PATH_INFO.  Usually gitweb sets its value correctly,
 	and there is no need to set this variable, e.g. to $my_uri or "/".
 	See `$per_request_config` if you need to override it anyway.
++
+You would need to set this variable when using path_info-based URLs
+while using gitweb as a directory handler (which means that full path
+to browse repositories looks like `http://git.example.com/gitweb`
+rather than looking like `http://git.example.com/gitweb/gitweb.cgi`).
+You can find yourself in this situation (gitweb as directory handler)
+when configuring gitweb to use FastCGI interface as shown in
+"Webserver configuration" section on linkgit:gitweb[1] manpage.
++
+If static files are served from 'static' subdirectory of directory
+the gitweb script is handler for, with default URLs of static files
+(e.g. `$logo` is `static/git-logo.png`), then you would need to ensure
+that `$base_url` ends with slash to denote that it is directory, to
+work correctly:
++
+----------------------------------------------------------------------
+$base_url .= '/' unless ($base_url =~ m!/$!);
+----------------------------------------------------------------------
++
+For example if gitweb URL is `http://git.example.com/gitweb`, and
+static files are available in `http://git.example.com/gitweb/static/`
+then `$base_url` must end up to be `http://git.example.com/gitweb/`
+(with trailing slash) for e.g. `static/git-logo.png` relative link
+to refer to `http://git.example.com/gitweb/static/git-logo.png`.
 
 
 CONFIGURING GITWEB FEATURES

diff --git a/Documentation/gitweb.txt b/Documentation/gitweb.txt
index b394ecc..c3db66a 100644
--- a/Documentation/gitweb.txt
+++ b/Documentation/gitweb.txt

@@ -473,6 +473,10 @@ With that configuration the full path to browse repositories would be:
 
   http://server/gitweb
 
+Note that for this configuration `$base_url` must be set as described
+in linkgit:gitweb.conf[5] for gitweb to correctly serve static files
+with path_info links.
+
 As PSGI using plackup
 ~~~~~~~~~~~~~~~~~~~~~
 Gitweb can run as PSGI app (via emulation with *CGI::Emulate::PSGI*(3pm)).

-- 
1.7.9

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hello again,

On 17/04/12 14:30, Jakub Narebski wrote:

On Mon, 16 Apr 2012, José María Escartín Esteban wrote:

quoted

On 14/04/12 19:19, Jakub Narebski wrote:

quoted

On Thu, 5 Apr 2012, Jakub Narebski wrote:

quoted

On Thu, 5 Apr 2012, José María Escartín Esteban wrote:

quoted

quoted

quoted

There might be problem if you configured your web server to serve
gitweb using it as a handler for subdirectory, so the script name
does not need to appear in URL, e.g.

http://localhost/cgi-bin/gitweb

which would require the following base element

<base href="http://localhost/cgi-bin/gitweb/" />

[...]

Yes, that was precisely the problem.

Note that you won't have problem with `http://localhost` as base url 
(without any directory), or in more realistic case using virtual host 
rather than virtual directory.  The problem is that without '/' the path
part of $base_url is stripped of last element.

Just for completeness.

Right.

[...]

quoted

This addition to the man page would of course have been useful, since it
took me a while to detect that the problem was a missing slash in the
base url. However, it would not have solved completely my problem, and
for me the best has been to add the line of code you proposed in the
previous email.

So how about the proposed updated addition at the end of this email?

I think now it's more clear and explicit, and more useful for users not
familiar with Perl, so better.  Thank you!

quoted

I am running a Debian testing (wheezy) server, so the distro gitweb files
are:

/etc/apache2/conf.d/gitweb /etc/gitweb.conf /usr/share/gitweb/gitweb.cgi 
/usr/share/gitweb/index.cgi -> /usr/share/gitweb/gitweb.cgi 
/usr/share/gitweb/static/*

I am using two sets of git repositories, and each of the sets works
through a particular gitolite user:

user     home ---------------------- git      /srv/project mygit
/srv/mygit

I wanted that different groups of people were able to browse the
'project' repos (in fact it is a mirror from the main project server)
from http://server.example.com/project/ , and the 'mygit' repos from 
http://server.example.com/mygit/ . And I wanted to keep urls as short and
easy as possible, and to avoid proliferation of configs as much as
possible.

I think that is this "two sets" that created configuration with a problem.

quoted

The setup I came up with was to link

/srv/www/project -> /usr/share/gitweb /srv/www/mygit ->
/usr/share/gitweb

and to use the following configuration files:

######### /etc/apache2/conf.d/gitweb ################## <Directory
/srv/www/project> SetEnv  GITWEB_PROJECTROOT      /srv/git/ SetEnv
GITWEB_CONFIG           /etc/gitweb.conf

[...]

quoted

#######################################################

######### /etc/gitweb.conf ############################ $projectroot =
$ENV{'GITWEB_PROJECTROOT'}."repositories"; $git_temp = "/tmp"; 
$projects_list = $ENV{'GITWEB_PROJECTROOT'}."projects.list"; @diff_opts =
(); $feature{'pathinfo'}{'default'} = [1]; 
$feature{'highlight'}{'default'} = [1]; $projects_list_description_width
= 50; #######################################################

A question: why not have

SetEnv  GITWEB_PROJECTROOT      /srv/git

and use

$projectroot = "$ENV{'GITWEB_PROJECTROOT'}/repositories"; $projects_list =
"$ENV{'GITWEB_PROJECTROOT'}/projects.list";

Aren't both configurations equivalent? (just moving a slash to one side or the
other of the definitions)

BTW. what is this $git_temp?  Is it modified gitweb, because core one 
doesn't need to use temporary directory for anything nowadays?

That option is enabled by default in all the Debian shipped /etc/gitweb.conf.
 It was like this the first time I installed the package from the Debian
repos, and it is still like this in the last version of the package in the
repos, gitweb_1.7.10-1_all.deb.  I didn't know it was any sort of modified
gitweb, the script is in git_1.7.10-1_all.deb (yes, they make a separate
package for gitweb but the script is in the git one), and the only differences
to the v1.7.10 upstream source seem to be installation variables/flags.

A tip: you can use GITWEB_CONFIG_COMMON for the common part of 
configuration and separate GITWEB_CONFIG for per-instance configuration. 
Though I am not sure if it would help in your case.  It requires modern
gitweb. though.

I will check that, thanks again.

Greetings,

	E.

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.12 (GNU/Linux)

iEYEARECAAYFAk+WT9YACgkQY+7weQMem3xchQCcCxsIOs7GOcluIb++KKweK76Y
TRcAnjauwtudLCc2qTwJd5D8FxwI34mC
=MbaI
-----END PGP SIGNATURE-----