Bug 31066 – [GTK] opening local files on win32

• Component: WebKitGTK
• Status: RESOLVED FIXED

[Josip]

Created attachment 42391 [details]
change setting of mime types on win os

Hi!
Due to mime-share-info misconfiguration or missing dependencies on user side local files ("file://") can not be opened. 
As g_file_info_get_content_type() returns just file extension (like ".html") and win32 use those to determine mime type can we use g_content_type_get_mime_type() on them like in attachment.

[Eric Seidel (no email)]

Comment on attachment 42391 [details]
change setting of mime types on win os

This patch needs a ChangeLog.  Also I don't think josip is a reviewer.  See http://webkit.org/coding/commit-review-policy.html

[Eric Seidel (no email)]

See http://webkit.org/coding/contributing.html for instructions on how to post patches to webkit.

[Josip]

Created attachment 42514 [details]
change setting of mime types on win os (2. try)

Sorry for confusion!
Added ChangeLog to patch (should read instructions more carefully).
Misunderstood "add" for "+" but why i was allowed to do that in first place, shouldn't that be restricted based on user status.
(Tool-tips for all options shows: Request for patch review)

[Martin Robinson]

I wonder if this might be the right thing to do for all platforms and not just for PLATFORM_WIN32.

[Holger Freyther]

(In reply to comment #0)
> Created an attachment (id=42391) [details]
> change setting of mime types on win os
> 
> Hi!
> Due to mime-share-info misconfiguration or missing dependencies on user side
> local files ("file://") can not be opened. 
> As g_file_info_get_content_type() returns just file extension (like ".html")
> and win32 use those to determine mime type can we use
> g_content_type_get_mime_type() on them like in attachment.

Why don't just fix the misconfiguration? I don't think this workaround is clever...

[Josip]

> 
> Why don't just fix the misconfiguration? I don't think this workaround is
> clever...

How to do that?
I make mime-shared-info for win32, make environment variables like in linux and run update-mime-database but "content_type" is still just the file suffix (".html" not "text/html"). What do i do wrong? Did i miss some dependencies? Is this working for someone and how?

[Holger Freyther]

I have little idea from Windows32. In general I would like to see a test case for Glib/Gio and then an explanation of why it is failing on windows. Once we have that we can decide on a fix?

[Josip]

(In reply to comment #7)
> I have little idea from Windows32. In general I would like to see a test case
> for Glib/Gio and then an explanation of why it is failing on windows. Once we
> have that we can decide on a fix?

Not that i know anything about that but maybe valid explanation is on: http://library.gnome.org/devel/gio/stable/gio-GContentType.html which says:

"""
A content type is a platform specific string that defines the type of a file. On unix it is a mime type, on win32 it is an extension string like ".doc", ".txt" or a percieved string like "audio". Such strings can be looked up in the registry at HKEY_CLASSES_ROOT.
"""

Not sure what you mean by test case, is that something for "run-webkit-tests" or this would do:

///////////////////////////////////////////////////////////////////////////////
/* gio_test */

#include <gio/gio.h>
#include <stdio.h>

int main(int argc, const char * argv[])
{
    g_type_init ();
    GFile * file = g_file_new_for_commandline_arg (argv[1]);
    GError *error = 0;
    GFileInfo* info = g_file_query_info (file, "standard::*", 
        G_FILE_QUERY_INFO_NONE , NULL, NULL);

    printf("If GFile is '%s' then from GFileInfo content type is '%s'\n", 
        g_file_get_basename(file), g_file_info_get_content_type(info));
    printf("and from MIME is '%s', ", 
        g_content_type_from_mime_type(g_file_info_get_content_type(info)));
    printf("even if his mime type is '%s'\n", 
    g_content_type_get_mime_type(g_file_info_get_content_type(info)));
	
    return (0);
}
///////////////////////////////////////////////////////////////////////////////

$ ./gio_test ~/wtest.html
If GFile is 'wtest.html' then from GFileInfo content type is 'text/html'
and from MIME is 'text/html', even if his mime type is 'text/html'

$ wine ./gio_test.exe ~/wtest.html 
If GFile is 'wtest.html' then from GFileInfo content type is '.html'
and from MIME is '(null)', even if his mime type is 'text/html'

[Martin Robinson]

There are some tests in WebKit/gtk/tests. You could probably make one much like the code you just posted.

In light of your previous comment, I'd say this fix does not need to be protected by a PLATFORM #ifdefs. It should apply equally to all platforms.

[Jan Alonzo]

Comment on attachment 42514 [details]
change setting of mime types on win os (2. try)

> +    #if PLATFORM(WIN_OS)
> +    response.setMimeType(g_content_type_get_mime_type(g_file_info_get_content_type(info)));
> +    #else
>      response.setMimeType(g_file_info_get_content_type(info));
> +    #endif

Were there any issues running the tests in WebKit/gtk/tests? You should be able to add a test for this issue in one of the unit tests in that directory.

Also, there shouldn't be a space before #if/#else/#endif. 

r- because of the extra spaces. It would be nice if we have test for this too.

[Josip]

All 15 tests from WebKit/gtk/tests passed OK

###############################################################################
C:\WebKitGTK-1.1.16\bin\unittests>FOR %f IN (*.exe) DO %f

C:\WebKitGTK-1.1.16\bin\unittests>testatk.exe
/webkit/atk/get_text_at_offset: OK

C:\WebKitGTK-1.1.16\bin\unittests>testdownload.exe
/webkit/download/create: OK
/webkit/download/perform: OK

C:\WebKitGTK-1.1.16\bin\unittests>testhittestresult.exe
/webkit/hittestresult/document: OK
/webkit/hittestresult/image: OK
/webkit/hittestresult/editable: OK
/webkit/hittestresult/link: OK

C:\WebKitGTK-1.1.16\bin\unittests>testhttpbackend.exe
/webkit/soupmessage/lifetime: OK

C:\WebKitGTK-1.1.16\bin\unittests>testkeyevents.exe
/webkit/keyevent/textfield: OK
/webkit/keyevent/buttons: OK
/webkit/keyevent/link: OK

C:\WebKitGTK-1.1.16\bin\unittests>testloading.exe
/webkit/loading/status: OK
/webkit/loading/error: OK
/webkit/loading/cancelled: OK

C:\WebKitGTK-1.1.16\bin\unittests>testnetworkrequest.exe
/webkit/networkrequest/createdestroy: OK
/webkit/networkrequest/properties: OK

C:\WebKitGTK-1.1.16\bin\unittests>testnetworkresponse.exe
/webkit/networkresponse/createdestroy: OK
/webkit/networkresponse/properties: OK

C:\WebKitGTK-1.1.16\bin\unittests>testwebbackforwardlist.exe
/webkit/webbackforwardlist/add_item: OK
/webkit/webbackforwardlist/list_order: OK
/webkit/webhistoryitem/lifetime: OK

C:\WebKitGTK-1.1.16\bin\unittests>testwebdatasource.exe
/webkit/webdatasource/get_initial_request: OK
/webkit/webdatasource/api: OK
/webkit/webdatasource/unreachable_uri: OK

C:\WebKitGTK-1.1.16\bin\unittests>testwebframe.exe
/webkit/webview/create_destroy: OK
/webkit/webframe/lifetime: OK
/webkit/webview/printing: OK

C:\WebKitGTK-1.1.16\bin\unittests>testwebhistoryitem.exe
/webkit/webhistoryitem/get_data: OK
/webkit/webhistoryitem/alternate_title: OK

C:\WebKitGTK-1.1.16\bin\unittests>testwebresource.exe
/webkit/webresource/get_url: OK
/webkit/webresource/get_mime_type: OK
/webkit/webresource/get_text_encoding_name: OK
/webkit/webresource/get_frame_name: OK
/webkit/webresource/get_data: OK
/webkit/webresource/loading: OK
/webkit/webresource/sub_resource_loading: OK

C:\WebKitGTK-1.1.16\bin\unittests>testwebsettings.exe
/webkit/websettings/user_agent: OK

C:\WebKitGTK-1.1.16\bin\unittests>testwindow.exe
/webkit/window/scrollbar_policy: OK
###############################################################################

same result with both patched and unpatched version as "testwebresorce" dont use "file:" but "http:". Is that what i should expand for testing local files.

As for spaces, it took me long to figure out what you meen :) Should check some surrounding code first.

[Martin Robinson]

Created attachment 66813 [details]
Patch for this issue with tests

[Gustavo Noronha (kov)]

Comment on attachment 66813 [details]
Patch for this issue with tests

Very good, looks like we can uncomment those ogg tests now, too as we discussed on IRC =).

[Martin Robinson]

Committed r68313: <http://trac.webkit.org/changeset/68313>